jax-diffusion-models-pydata-boston-2025

A Beginner’s Friendly Introduction to Diffusion Models in JAX

Axel Donath, PyData Meeting Boston June 18th, 2025

Who Am I? I am a Postdoctoral Researcher at the Center for Astrophysics | Harvard & Smithsonian where I research in the field X-Ray and Gamma-Ray Astronomy. I develop new AI, machine learning and statistical methods as well as scientific open source software for the analysis of astronomical low counts data. I’m also editor for the Astronomy and Astrophysics track of the Journal of Open Source Software JOSS.

You can find out more about me and my work at:

This tutorial notebook is a beginner’s friendly introduction to diffusion models in JAX. It is mostly focused on introducing the core concepts of diffusion models from first principles along with teaching the basics of JAX, a differentiable, hardware accelerated, array library for deep learning and scientific computing.

Prerequisites

What are Diffusion Models?

Diffusion models are a class of generative models that are based on the concept of iteratively removing noise from an image or other data. “Generative” means they model a probability distribution and can be used to draw new samples from the distribution and thus generate new plausible images. Or an alternative way of thinking about diffusion models is to think of them as sampling algorithms, that sample from a fixed distribution. Where the distribution we sample from is defined by the distribution of the data we use for training them.

Image taken from: https://yang-song.net/blog/2021/score/

However sampling from high dimensional distributions, such as the distribution of natural images, where each pixel is treated as an independent dimension, can be technically quite challenging. Diffusion models are a way to make this sampling process feasible and scalable. Diffusion model are mathematically well founded and draw many parallels to actual physical diffusion processes, denoising methods in image processing as well as theory of (stochastic) differential equations.

In this tutorial We will take a look at a specific type of diffusion models called Denoising Diffusion Probabilistic Models (DDPMs) introduced by Ho et al. in 2020.. We will use simple “toy examples” on 1d and 2d data to illustrate and understand the method. The goal is to build intuition and provide a solid foundation for working with production scale models.

Note: I personally find that “diffusion models” are to a certain degree a misnomer, as the relevant part of the model is the reverse process, where the diffuson process is inverted. But we will learn about that later.

What is JAX?

In a single sentence you can think of JAX as Numpy with support for GPUs and automatic differentiation. It also includes a just in time compilation (JIT) and function transforms, such as vectorization (like np.vectorize) which is automatically lowered to hardware accelerated XLA (XLA is an open source compiler for linear algebra / machine learnin) instructions.

The JAX docs are available at: https://docs.jax.dev/en/latest/

There is diverse ecosystem of open source tools build around JAX covering:

There is also a curated list of resources around JAX here: https://github.com/n2cholas/awesome-jax

Quick Survey

More Ressources

An excellent intuitive introduction to the math behind diffusion models can be found in the following video series by Deepia on Youtube:

If you would like to prepare for this tutorial I would recommend to watch those videos in advance. They are each 15-30 minutes long and watching them in the given order is recommended.

If you are interested in more details and different perspectives about diffusion models I recommend to checkout the following blog posts:

Other minimal implementations in PyTorch:

Glossary

# uncomment on google collab to install dependencies
#import sys
#!{sys.executable} -m pip install -r https://raw.githubusercontent.com/adonath/jax-diffusion-models-pydata-boston-2025/main/requirements.txt
#!mkdir images

Warm Up: the Diffusion Process

Diffusion is typically a physical process where particles spread from regions of high concentration to regions of low concentration over time, driven by random motion. In nature, this can be seen in phenomena like the dispersal of ink in water or the flow of heat through a solid. Mathematically, diffusion is often modeled using stochastic processes or differential equations, capturing how randomness leads to gradual smoothing and homogenization of an initial state. In a more general understanding diffusion can also include convection, meaning a systematic drift of the particles driven by an external force, such as wind, a gravity driven flow in water, or stiring in a mixing process.

We will first explore the process of diffusion on a simple example in JAX. Let’s start with the relevant imports:

from functools import partial
from collections import namedtuple

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import animation

import jax
from jax import numpy as jnp
from jax import random

The jax.numpy namespace can be thought of as a “drop-in” replacement for Numpy. It was designed to be as similar as possible to the standard numpy namespace. In almost all cases you will find the equivalent of some np.<> function in jnp.<>. Howevere there is one important difference: array objects in JAX are not mutable, meaning they cannot be modified in place. Something like array[2:3] = 3, does not work in JAX, for reasons that become clear later, when we dicuss just in time compilation (JIT). There is a handful more of these differences which are documented in “JAX - The Sharp Bits”.

For this illustrative example we will work with a simple mixture of Gaussians as the inital distribution of the “particles” (I will use “particles” and “samples” synonymously). In JAX we can just use functionality in the jnp.<> namespace to define our model:

def gaussian(x, norm, mu, sigma):
    """Single Gaussian distribution"""
    return norm * jnp.exp(-0.5 * ((x - mu) / sigma) ** 2) / (sigma * jnp.sqrt(2 * np.pi))


def gmm(x, norm, mu, sigma):
    """Gaussian Mixture Model"""
    values = jnp.sum(gaussian(x, norm, mu, sigma), axis=0) / mu.shape[0]

    # later we compute the gradient, which requires a returning a scalar value
    if values.shape == (1,):
        return values[0]
    
    return values

norm, mu, sigma = jnp.array([1, 1])[:, None], jnp.array([-1, 1])[:, None], jnp.array([0.25, 0.25])[:, None]

x_plot = jnp.linspace(-2, 2, 1000)
y = gmm(x_plot, norm, mu, sigma)

This is what the model looks like:

ax = plt.subplot()
ax.plot(x_plot, y)
ax.set_xlabel('x')
ax.set_ylabel('p(x)')

Text(0, 0.5, 'p(x)')

png

Now we proceed to drawing samples from the distribution. As it is a Gaussian distribution we can directly sample from it using the build-in sampling methods from JAX. They are found in the jax.random namespace. This is again very much like numpy, however there is an important difference: instead of a global RandomState object JAX uses an approach, based on “keys”, which can be split locally and are passed to the random sampling method:

n_samples_ = 500_000

key = random.key(9834)

# TBD: this assumes equal norm, draw from bernoulli instead...
x_init = sigma * random.normal(key, (2, n_samples_ // 2,)) + mu

Let’s make a histogram of the samples to verify the shape of the distribution:

ax = plt.subplot()
ax.hist(x_init.flatten(), bins=100, density=True, histtype='step', label="Initial samples");
ax.set_xlabel('$x_i$')
ax.set_ylabel('$p(x_i)$');

png

Now we introduce the actual diffusion process. The diffusion process we choose here consists of two distinct steps or terms:

In addition we assume $\beta_t \in [0, 1] $ and $\beta_t \to 1$ with time. With this specific choice of can achieve that the process finally converges towards a normal Gaussian distribution $\mathcal{N}(0, 1)$. The choice of the factors given above seems a bit “magic” here, but the reasoning is given in more detail in the DDPM paper I linked above (hint: in addition to leading to convergence to a unit Gaussian it also allows for computing the n-th time step in closed form, which will see later in the tutorial).

For now we can just numerically convince ourselves that this process indeed transforms our two component GMM into a normal Gaussian Distribution:

n_iter = 100
beta_t = jnp.linspace(0, 1, n_iter)
x = x_init.flatten()

sample_trace_diffusion_beta = []

for idx, beta in enumerate(beta_t):
    key, sub_key = random.split(key)
    x = jnp.sqrt(1. - beta) * x + beta * random.normal(key=sub_key, shape=x.shape)
    sample_trace_diffusion_beta.append(x)

sample_trace_diffusion_beta = jnp.stack(sample_trace_diffusion_beta, axis=0)

To visualize the process over time, we can make a histogram of the samples at each point in time t. Of course we can use a Python loop for this and call e.g. np.histogram and append all results, however this is a good time to introduce a first function transform in JAX, named jax.vmap. jax.vmap allows to map an arbitrary function over the leading dimension of an array. It is comparable to np.vectorize, however the latter literally implements a Python loop, while jax.vmap lowers the operation to compiled code and is thus much more performant.

So now we can use it to map jnp.histogram over the time dimension of our samples:

default_hist = partial(jnp.histogram, bins=100, range=(-3, 3), density=True)

batched_histogram = jax.vmap(default_hist)

For covenience we combined this into a little helper function, we can use to see how the density changes with time and plot the trajectory of some example particles:

def plot_trace(trace, n_traces=5, ax=None, x_min=-3, x_max=3):
    """Plot distribution at multiple points in time as trace"""
    hist_values, _ = batched_histogram(trace)    

    n_iter, n_samples = trace.shape

    ax = plt.subplot() or ax
    ax.imshow(hist_values.T[:, :], extent=[0, n_iter, x_min, x_max], aspect="auto", origin="lower")

    # plot some example traces
    key = random.PRNGKey(9823)
    for idx in random.randint(key, (n_traces,), 0, n_samples):
        ax.plot(trace[:, idx])

    ax.set_ylim(x_min, x_max)
    ax.set_xlabel("# Iteration")
    ax.set_ylabel("x")
    return ax

plot_trace(sample_trace_diffusion_beta);

png

We can easily convince ourself, that the final (stationary) distribution is indeed a standard Gaussian:

ax = plt.subplot()
ax.hist(sample_trace_diffusion_beta[-1], density=True, bins=50, label="Samples")

x_plot = jnp.linspace(-5, 5, 100)
ax.plot(x_plot, gaussian(x_plot, norm=1, mu=0, sigma=1), label="Unit Gaussian")
ax.set_xlabel("x")
ax.set_ylabel("PDF")
ax.legend();

png

“Reversing” Diffusion via Langevin Dynamics

Now that we have “destroyed” our intial distribution (the two component Gaussian) with the forward diffusion process, wouldn’t it be great if we could reverse the process? This would give us a sampling algorithm, that allowed us to sample from an arbitrary distribution. Starting from a normal Gaussian distribution we would transform the samples back into the shape of the intial distribution.

It turns out such an algorithm exists, it is named “Langevin Dynamics”. However the algorithm requires one additional piece of information: the score function of the target distribution. The score function is defined as the gradient of the log of the distribution.

In JAX we can directly get the score function by taking the gradient of the log using jax.grad():

def log_gmm(x, norm, mu, sigma):
    """Log of the GMM"""
    return jnp.log(gmm(x, norm, mu, sigma))

gmm_log_part = partial(log_gmm, norm=norm, mu=mu, sigma=sigma) # the partial just simplifies not parsing the parameters later...
score_fun = jax.vmap(jax.grad(gmm_log_part))

Now we can see what it looks like:

ax = plt.subplot()
x_plot = jnp.linspace(-2, 2, 1000)
ax.plot(x_plot, score_fun(x_plot))
ax.set_xlabel('x')
ax.set_ylabel('d/dx log p(x)');

png

Now let us implement a reverse diffusion process based on “Langevin Dynamics”. Starting from an initial sample $x_i$, the algorithm looks like this:

$\begin{align} \mathbf{x}{i+1} \gets \mathbf{x}_i + 0.5 \alpha \nabla\mathbf{x} \log p(\mathbf{x}) + \sqrt{\alpha}~ \mathbf{z}_i, \quad i=0,1,\cdots, K, \end{align}$

Where $\mathbf{z}i \sim \mathcal{N}(0, I)$. When $\alpha \to 0$ and $K \to \infty$, this converges to the target distribution, defined by its score function $\nabla\mathbf{x} \log p(\mathbf{x})$.

The intuition behind the algorithm is the following: give a starting point for a sample $x_0$ it is moved along the gradient (gradient descent!) to move towards the region of highest probability. However whith just this component all samples would end up in the same final location, the maximum of the distribution. To prevent this, the scale of the gradient is decreased with time ($\alpha$) and an addititional random update is added, which moves the sample in a random direction. The method was first introduced by Welling et al. 2011, where they also present proof of convergence.

While we could implement the algorithm the same way we did above using a Python loop, we make use of another JAX fundemantal, named jax.lax.scan. lax.scan allows you scan a function over the leading array axes while carrying along a state. It requires us to define the “body” of the loop in a dedicated function:

# JAX can natively handle "structs of arrays" or "PyTrees"
SampleArgs = namedtuple("Args", ["key", "idx", "x", "alpha_0", "p_0"])

def sample(score, args, _):
    # compute the "learning rate" depending on the iteration
    alpha = args.alpha_0 * (args.p_0 ** args.idx) ** 2

    # sample stochastic update
    key, subkey = random.split(args.key)
    dx = random.normal(subkey, args.x.shape)

    # combine the gradient and the stochastic update
    x = args.x + 0.5 * alpha * score(args.x) + jnp.sqrt(alpha) * dx
    return SampleArgs(key, args.idx + 1, x, args.alpha_0, args.p_0), x

Note on pure functions: Functions passed to jax.lax.scan are automatically JIT compiled, which allows for efficient execution. However this requires functions to be “pure” or alternatively they cannot have “side effects”. Side effects include e.g. modifying an object / data that was passed to the function as argument in place, printing something or interacting with global state in the function. This limitation becomes clear, when you remind yourself that for JIT compilation the Python code only serves as a laguage to define the computational graph and is never really executed. Instead a JIT compiled version of the code is executed, which is compiled to lower level hardware accelerated code. How should a remote GPU “know” how to modify a specific Python object in thh main process place? This would require supporting complex communication between process and platforms. The requirement of pure functions makes the implementation of the JIT much simpler and more stable.

Now we can use that function by defining the initial state first and calling jax.lax.scan:

n_samples = 100_000
n_iter = 500

key = random.PRNGKey(42)
key, subkey = random.split(key)

init = SampleArgs(
    key=key,
    idx=0,
    x=random.normal(subkey, (n_samples,)),
    alpha_0=0.002,
    p_0=1.0
)

result, sample_trace = jax.lax.scan(partial(sample, score_fun), init, length=n_iter)

plot_trace(sample_trace);

png

ax = plt.subplot()
ax.hist(sample_trace[-1], density=True, bins=70, label="Samples")

x_plot = jnp.linspace(-2, 2, 100)
y = gmm(x_plot, norm, mu, sigma)
ax.plot(x_plot, y, label="Target distribution")
ax.set_xlim()
ax.set_xlabel("x")
ax.set_ylabel("PDF")
ax.legend();

png

So we have transformed the samples from a unit Gaussian into the target distribution using Langevin dynamics!

Score Matching

In this next section we will go from the simple Langevin dynamics example to Denoising Diffusion Probabilistic Models (DDPM) Ho et al. The Langevin dynamics example required the knowledge of the score function, which we derived from the bimodal Gaussian PDF using jax.grad.

However in general we do not have access to the score function, we just have access to the data distribution. Still we can estimate the score function from the data in a process called “score matching”, introduced by Hyvarinen, 2005. It is intuitive that to “match the score” we could minimize the expectation value of the squared difference between a model function $\mathcal{F}_{\theta}(\mathbf{x})$ and the score function computed from the data distribution (leaving open how to do this for now):

\[\mathcal{L}_{mse} = E_{\mathbf{x} \sim p(\mathbf{x})} \left[ \left\lVert \mathcal{F}_{\theta}(\mathbf{x}) - \nabla_{\mathbf{x}} \log p(\mathbf{x}) \right\lVert_2^2 \right]\]

The core of the method of “score matching” involves the intesting result, that the MSE loss between the score function and the gradient of the log PDF can be reformulated as:

\[\mathcal{L}_{matching} = E_{\mathbf{x} \sim p(\mathbf{x})} \left[ \text{ tr}\left( \nabla_{\mathbf{x}} \mathcal{F}_{\theta}(\mathbf{x}) \right) + \frac{1}{2} \left\Vert \mathcal{F}_{\theta}(\mathbf{x}) \right\lVert_2^2 \right]\]

If you are interested in the mathematical derivation of this new loss, you can checkout e.g. http://yang-song.net/blog/2019/ssm/.

from sklearn.datasets import make_swiss_roll
import equinox as eqx
import optax

For this next section we will work with another toy dataset, the “swiss roll”:

random_state = np.random.RandomState(8923)

n_samples = 1_000
x = make_swiss_roll(n_samples=n_samples, noise=0.1, random_state=random_state)[0]
x = 0.13 * x[:, [0, 2]] # drop z axis

This is what the dataset look like:

ax = plt.subplot()

ax.scatter(x[:, 0], x[:, 1], s=1)
ax.set_aspect('equal')

scale = 2.0
ax.set_xlim(-scale, scale)
ax.set_ylim(-scale, scale)
ax.set_xlabel("x")
ax.set_ylabel("y");

png

To model the score function we choose a simple multi-layer perceptron. In JAX there are multiple choices of deep learning libraries. I personally found Equinox easiest to work with. However if you come from PyTorch than flax / nnx might be even closer.

The model definition in equinox looks like:

class MLP(eqx.Module):
    """Simple MLP"""
    fc1: eqx.nn.Linear
    fc2: eqx.nn.Linear
    fc3: eqx.nn.Linear

    def __init__(self, in_dim, out_dim, key, hidden_dim=128):
        keys = jax.random.split(key, 3)
        self.fc1 = eqx.nn.Linear(in_dim, hidden_dim, key=keys[0])
        self.fc2 = eqx.nn.Linear(hidden_dim, hidden_dim, key=keys[1])
        self.fc3 = eqx.nn.Linear(hidden_dim, out_dim, key=keys[2])

    def __call__(self, x):
        x = jax.nn.gelu(self.fc1(x))
        x = jax.nn.gelu(self.fc2(x))
        x = self.fc3(x)
        return x

I you have worked with PyTorch before, this should still look familiar to you. Models are defined by inhereting from a Module object and implementing the forward pass, here defined in the __call__ method. New are the class level declarations of class members, which are “borrowed” from dataclasses and the splitting of the random state, we have seen before.

One particular design choice in Equinox is that you never have to handle the “batch axis”, but you can always use jax.vmap over an entire model definition. This is possible because eqx.Module objects are “structs of arrays” supported by JAX. However one thing to keep in mind (a “sharp bit”) is that these structures can only be trees and not graphs! This makes for example sharing parameters a bit more complex.

Next we implement the loss function I we have defined above and to make the function more efficient, we will use JAX’s built-in “just in time” compilation. The way this works is by decorating the function definition with the jax.jit object.

@jax.jit
def score_matching_loss(model, samples):
    """Score matching loss"""
    logp = jax.vmap(model)(samples)
    # Compute the norm loss
    norm_loss = jnp.linalg.norm(logp, axis=-1) ** 2 / 2.0
    
    # Compute the Jacobian loss
    jacob_mat = jax.vmap(jax.jacobian(model))(samples)
    tr_jacobian_loss = jnp.trace(jacob_mat, axis1=-2, axis2=-1)
    
    return (tr_jacobian_loss + norm_loss).mean()

Next we train the model on the dataset, using the Adam optimzer and the optax library. Equinox provide some extra methods such as eqx.filter_jit, which can be thought of an convenience wrapper around the jax.jit decorator, which automacilly filters which attributes of the model are arrays and relevant for JIT and gradient computation. I will not go into forther explanation, but refer to the Equinox docs instead.

model = MLP(2, 2, key=jax.random.key(1234), hidden_dim=128)

optimizer = optax.adam(learning_rate=1e-3)
opt_state = optimizer.init(model)

n_iter = 1000

@eqx.filter_jit
def make_step(model, opt_state, x):
    """Make a single step in the optimization routine"""
    loss_value, grads = eqx.filter_value_and_grad(score_matching_loss)(model, x)
    updates, opt_state = optimizer.update(
        grads, opt_state, eqx.filter(model, eqx.is_array)
    )
    model = eqx.apply_updates(model, updates)
    return model, opt_state, loss_value


for idx in range(n_iter):
    model, opt_state, train_loss = make_step(model, opt_state, x)
    print(f"Iter {idx}, loss: {train_loss:.4f}", end="\r")

Iter 0, loss: 0.0227 Iter 1, loss: -0.0900 Iter 2, loss: -0.1967 Iter 3, loss: -0.2977 Iter 4, loss: -0.3934 Iter 5, loss: -0.4842 Iter 6, loss: -0.5703 Iter 7, loss: -0.6518 Iter 8, loss: -0.7290 Iter 9, loss: -0.8016 Iter 10, loss: -0.8695 Iter 11, loss: -0.9327 Iter 12, loss: -0.9906 Iter 13, loss: -1.0432 Iter 14, loss: -1.0901 Iter 15, loss: -1.1312 Iter 16, loss: -1.1667 Iter 17, loss: -1.1966 Iter 18, loss: -1.2216 Iter 19, loss: -1.2421 Iter 20, loss: -1.2591 Iter 21, loss: -1.2731 Iter 22, loss: -1.2851 Iter 23, loss: -1.2958 Iter 24, loss: -1.3060

Iter 25, loss: -1.3161 Iter 26, loss: -1.3268 Iter 27, loss: -1.3382 Iter 28, loss: -1.3507 Iter 29, loss: -1.3642 Iter 30, loss: -1.3785 Iter 31, loss: -1.3936 Iter 32, loss: -1.4091 Iter 33, loss: -1.4248 Iter 34, loss: -1.4404 Iter 35, loss: -1.4558 Iter 36, loss: -1.4709 Iter 37, loss: -1.4857 Iter 38, loss: -1.5002 Iter 39, loss: -1.5145 Iter 40, loss: -1.5287 Iter 41, loss: -1.5429 Iter 42, loss: -1.5571 Iter 43, loss: -1.5713 Iter 44, loss: -1.5857 Iter 45, loss: -1.6002 Iter 46, loss: -1.6148 Iter 47, loss: -1.6296 Iter 48, loss: -1.6445 Iter 49, loss: -1.6597

Iter 50, loss: -1.6750 Iter 51, loss: -1.6907 Iter 52, loss: -1.7067 Iter 53, loss: -1.7230 Iter 54, loss: -1.7398 Iter 55, loss: -1.7569 Iter 56, loss: -1.7745 Iter 57, loss: -1.7925 Iter 58, loss: -1.8109 Iter 59, loss: -1.8299 Iter 60, loss: -1.8493 Iter 61, loss: -1.8693 Iter 62, loss: -1.8899 Iter 63, loss: -1.9111 Iter 64, loss: -1.9329 Iter 65, loss: -1.9555 Iter 66, loss: -1.9787 Iter 67, loss: -2.0028 Iter 68, loss: -2.0278 Iter 69, loss: -2.0537 Iter 70, loss: -2.0807 Iter 71, loss: -2.1086 Iter 72, loss: -2.1377 Iter 73, loss: -2.1680 Iter 74, loss: -2.1995

Iter 75, loss: -2.2322 Iter 76, loss: -2.2662 Iter 77, loss: -2.3016 Iter 78, loss: -2.3384 Iter 79, loss: -2.3767 Iter 80, loss: -2.4165 Iter 81, loss: -2.4579 Iter 82, loss: -2.5010 Iter 83, loss: -2.5458 Iter 84, loss: -2.5924 Iter 85, loss: -2.6410 Iter 86, loss: -2.6914 Iter 87, loss: -2.7438 Iter 88, loss: -2.7982 Iter 89, loss: -2.8547 Iter 90, loss: -2.9134 Iter 91, loss: -2.9742 Iter 92, loss: -3.0373 Iter 93, loss: -3.1027 Iter 94, loss: -3.1704 Iter 95, loss: -3.2405 Iter 96, loss: -3.3130 Iter 97, loss: -3.3881 Iter 98, loss: -3.4658 Iter 99, loss: -3.5460 Iter 100, loss: -3.6290

Iter 101, loss: -3.7147 Iter 102, loss: -3.8032 Iter 103, loss: -3.8946 Iter 104, loss: -3.9890 Iter 105, loss: -4.0865 Iter 106, loss: -4.1871 Iter 107, loss: -4.2910 Iter 108, loss: -4.3982 Iter 109, loss: -4.5091 Iter 110, loss: -4.6235 Iter 111, loss: -4.7418 Iter 112, loss: -4.8641 Iter 113, loss: -4.9905 Iter 114, loss: -5.1212 Iter 115, loss: -5.2563 Iter 116, loss: -5.3962 Iter 117, loss: -5.5409 Iter 118, loss: -5.6906 Iter 119, loss: -5.8456 Iter 120, loss: -6.0061 Iter 121, loss: -6.1722 Iter 122, loss: -6.3441 Iter 123, loss: -6.5222 Iter 124, loss: -6.7065 Iter 125, loss: -6.8973

Iter 126, loss: -7.0949 Iter 127, loss: -7.2994 Iter 128, loss: -7.5111 Iter 129, loss: -7.7302 Iter 130, loss: -7.9569 Iter 131, loss: -8.1916 Iter 132, loss: -8.4343 Iter 133, loss: -8.6854 Iter 134, loss: -8.9451 Iter 135, loss: -9.2136 Iter 136, loss: -9.4911 Iter 137, loss: -9.7779 Iter 138, loss: -10.0743 Iter 139, loss: -10.3803 Iter 140, loss: -10.6963 Iter 141, loss: -11.0223 Iter 142, loss: -11.3586 Iter 143, loss: -11.7054 Iter 144, loss: -12.0628 Iter 145, loss: -12.4309 Iter 146, loss: -12.8100 Iter 147, loss: -13.2000 Iter 148, loss: -13.6013 Iter 149, loss: -14.0139 Iter 150, loss: -14.4379 Iter 151, loss: -14.8734

Iter 152, loss: -15.3207 Iter 153, loss: -15.7797 Iter 154, loss: -16.2506 Iter 155, loss: -16.7336 Iter 156, loss: -17.2287 Iter 157, loss: -17.7361 Iter 158, loss: -18.2559 Iter 159, loss: -18.7883 Iter 160, loss: -19.3333 Iter 161, loss: -19.8911 Iter 162, loss: -20.4618 Iter 163, loss: -21.0456 Iter 164, loss: -21.6426 Iter 165, loss: -22.2529 Iter 166, loss: -22.8767 Iter 167, loss: -23.5140 Iter 168, loss: -24.1651 Iter 169, loss: -24.8300 Iter 170, loss: -25.5088 Iter 171, loss: -26.2018 Iter 172, loss: -26.9090 Iter 173, loss: -27.6306 Iter 174, loss: -28.3668 Iter 175, loss: -29.1175 Iter 176, loss: -29.8830 Iter 177, loss: -30.6634

Iter 178, loss: -31.4587 Iter 179, loss: -32.2691 Iter 180, loss: -33.0946 Iter 181, loss: -33.9351 Iter 182, loss: -34.7908 Iter 183, loss: -35.6615 Iter 184, loss: -36.5473 Iter 185, loss: -37.4480 Iter 186, loss: -38.3638 Iter 187, loss: -39.2946 Iter 188, loss: -40.2404 Iter 189, loss: -41.2013 Iter 190, loss: -42.1774 Iter 191, loss: -43.1689 Iter 192, loss: -44.1761 Iter 193, loss: -45.1991 Iter 194, loss: -46.2382 Iter 195, loss: -47.2938 Iter 196, loss: -48.3660 Iter 197, loss: -49.4552 Iter 198, loss: -50.5615 Iter 199, loss: -51.6850 Iter 200, loss: -52.8258 Iter 201, loss: -53.9838 Iter 202, loss: -55.1586 Iter 203, loss: -56.3502

Iter 204, loss: -57.5580 Iter 205, loss: -58.7818 Iter 206, loss: -60.0213 Iter 207, loss: -61.2764 Iter 208, loss: -62.5473 Iter 209, loss: -63.8344 Iter 210, loss: -65.1382 Iter 211, loss: -66.4592 Iter 212, loss: -67.7978 Iter 213, loss: -69.1542 Iter 214, loss: -70.5283 Iter 215, loss: -71.9201 Iter 216, loss: -73.3295 Iter 217, loss: -74.7562 Iter 218, loss: -76.2002 Iter 219, loss: -77.6613 Iter 220, loss: -79.1395 Iter 221, loss: -80.6349 Iter 222, loss: -82.1476 Iter 223, loss: -83.6776 Iter 224, loss: -85.2250 Iter 225, loss: -86.7898 Iter 226, loss: -88.3721 Iter 227, loss: -89.9714 Iter 228, loss: -91.5858 Iter 229, loss: -93.2189

Iter 230, loss: -94.8697 Iter 231, loss: -96.5346 Iter 232, loss: -98.2172 Iter 233, loss: -99.9183 Iter 234, loss: -101.6335 Iter 235, loss: -103.3678 Iter 236, loss: -105.1203 Iter 237, loss: -106.8878 Iter 238, loss: -108.6752 Iter 239, loss: -110.4806 Iter 240, loss: -112.3021 Iter 241, loss: -114.1434 Iter 242, loss: -116.0030 Iter 243, loss: -117.8794 Iter 244, loss: -119.7756 Iter 245, loss: -121.6905 Iter 246, loss: -123.6231 Iter 247, loss: -125.5753 Iter 248, loss: -127.5469 Iter 249, loss: -129.5367 Iter 250, loss: -131.5455 Iter 251, loss: -133.5738 Iter 252, loss: -135.6208 Iter 253, loss: -137.6861 Iter 254, loss: -139.7699 Iter 255, loss: -141.8723

Iter 256, loss: -143.9931 Iter 257, loss: -146.1316 Iter 258, loss: -148.2876 Iter 259, loss: -150.4610 Iter 260, loss: -152.6514 Iter 261, loss: -154.8570 Iter 262, loss: -157.0769 Iter 263, loss: -159.3072 Iter 264, loss: -161.5640 Iter 265, loss: -163.8353 Iter 266, loss: -166.1150 Iter 267, loss: -168.4131 Iter 268, loss: -170.7330 Iter 269, loss: -173.0684 Iter 270, loss: -175.4179 Iter 271, loss: -177.7889 Iter 272, loss: -180.1821 Iter 273, loss: -182.5898 Iter 274, loss: -185.0142 Iter 275, loss: -187.4615 Iter 276, loss: -189.9276 Iter 277, loss: -192.4092 Iter 278, loss: -194.9090 Iter 279, loss: -197.4271 Iter 280, loss: -199.9622 Iter 281, loss: -202.5085

Iter 282, loss: -205.0715 Iter 283, loss: -207.6566 Iter 284, loss: -210.2579 Iter 285, loss: -212.8668 Iter 286, loss: -215.4897 Iter 287, loss: -218.1354 Iter 288, loss: -220.8017 Iter 289, loss: -223.4804 Iter 290, loss: -226.1716 Iter 291, loss: -228.8825 Iter 292, loss: -231.6142 Iter 293, loss: -234.3628 Iter 294, loss: -237.1254 Iter 295, loss: -239.9042 Iter 296, loss: -242.7020 Iter 297, loss: -245.5190 Iter 298, loss: -248.3518 Iter 299, loss: -251.1974 Iter 300, loss: -254.0504 Iter 301, loss: -256.9091 Iter 302, loss: -259.7653 Iter 303, loss: -262.6561 Iter 304, loss: -265.5446 Iter 305, loss: -268.4345 Iter 306, loss: -271.3401 Iter 307, loss: -274.2614

Iter 308, loss: -277.1756 Iter 309, loss: -280.0966 Iter 310, loss: -283.0561 Iter 311, loss: -286.0235 Iter 312, loss: -288.9783 Iter 313, loss: -291.9572 Iter 314, loss: -294.9661 Iter 315, loss: -297.9736 Iter 316, loss: -300.9843 Iter 317, loss: -304.0215 Iter 318, loss: -307.0739 Iter 319, loss: -310.1261 Iter 320, loss: -313.1921 Iter 321, loss: -316.2799 Iter 322, loss: -319.3800 Iter 323, loss: -322.4842 Iter 324, loss: -325.6017 Iter 325, loss: -328.7382 Iter 326, loss: -331.8887 Iter 327, loss: -335.0462 Iter 328, loss: -338.2126 Iter 329, loss: -341.3889 Iter 330, loss: -344.5747 Iter 331, loss: -347.7515 Iter 332, loss: -350.9250 Iter 333, loss: -354.0678

Iter 334, loss: -357.2769 Iter 335, loss: -360.5068 Iter 336, loss: -363.7079 Iter 337, loss: -366.8625 Iter 338, loss: -370.0291 Iter 339, loss: -373.2545 Iter 340, loss: -376.4775 Iter 341, loss: -379.6800 Iter 342, loss: -382.8868 Iter 343, loss: -386.1098 Iter 344, loss: -389.3420 Iter 345, loss: -392.5762 Iter 346, loss: -395.8273 Iter 347, loss: -399.0829 Iter 348, loss: -402.3310 Iter 349, loss: -405.5853 Iter 350, loss: -408.8578 Iter 351, loss: -412.1476 Iter 352, loss: -415.4341 Iter 353, loss: -418.7157 Iter 354, loss: -422.0053 Iter 355, loss: -425.3081 Iter 356, loss: -428.6200 Iter 357, loss: -431.9312 Iter 358, loss: -435.2425 Iter 359, loss: -438.5572

Iter 360, loss: -441.8821 Iter 361, loss: -445.2120 Iter 362, loss: -448.5436 Iter 363, loss: -451.8642 Iter 364, loss: -455.1793 Iter 365, loss: -458.4778 Iter 366, loss: -461.7929 Iter 367, loss: -465.1023 Iter 368, loss: -468.4242 Iter 369, loss: -471.7304 Iter 370, loss: -475.0366 Iter 371, loss: -478.3496 Iter 372, loss: -481.6644 Iter 373, loss: -484.9704 Iter 374, loss: -488.2654 Iter 375, loss: -491.5669 Iter 376, loss: -494.8855 Iter 377, loss: -498.2295 Iter 378, loss: -501.5741 Iter 379, loss: -504.9099 Iter 380, loss: -508.2411 Iter 381, loss: -511.5859 Iter 382, loss: -514.9460 Iter 383, loss: -518.3112 Iter 384, loss: -521.6710 Iter 385, loss: -525.0239

Iter 386, loss: -528.3748 Iter 387, loss: -531.7237 Iter 388, loss: -535.0691 Iter 389, loss: -538.3907 Iter 390, loss: -541.6752 Iter 391, loss: -544.8625 Iter 392, loss: -548.0308 Iter 393, loss: -551.2487 Iter 394, loss: -554.6914 Iter 395, loss: -557.9983 Iter 396, loss: -561.0600 Iter 397, loss: -564.1713 Iter 398, loss: -567.4440 Iter 399, loss: -570.6678 Iter 400, loss: -573.7614 Iter 401, loss: -576.9175 Iter 402, loss: -580.1248 Iter 403, loss: -583.2451 Iter 404, loss: -586.3752 Iter 405, loss: -589.5761 Iter 406, loss: -592.7174 Iter 407, loss: -595.8078 Iter 408, loss: -598.9728 Iter 409, loss: -602.1623 Iter 410, loss: -605.2737

Iter 411, loss: -608.3857 Iter 412, loss: -611.5595 Iter 413, loss: -614.7137 Iter 414, loss: -617.8305 Iter 415, loss: -620.9766 Iter 416, loss: -624.1452 Iter 417, loss: -627.2833 Iter 418, loss: -630.4088 Iter 419, loss: -633.5637 Iter 420, loss: -636.7290 Iter 421, loss: -639.8746 Iter 422, loss: -643.0142 Iter 423, loss: -646.1715 Iter 424, loss: -649.3369 Iter 425, loss: -652.4914 Iter 426, loss: -655.6395 Iter 427, loss: -658.7968 Iter 428, loss: -661.9650 Iter 429, loss: -665.1329 Iter 430, loss: -668.2941 Iter 431, loss: -671.4517 Iter 432, loss: -674.6078 Iter 433, loss: -677.7471 Iter 434, loss: -680.8379 Iter 435, loss: -683.7963

Iter 436, loss: -686.6616 Iter 437, loss: -689.5497 Iter 438, loss: -692.9543 Iter 439, loss: -696.3003 Iter 440, loss: -699.1670 Iter 441, loss: -701.9804 Iter 442, loss: -705.0797 Iter 443, loss: -708.2004 Iter 444, loss: -711.1052 Iter 445, loss: -714.0075 Iter 446, loss: -716.9669 Iter 447, loss: -719.8940 Iter 448, loss: -722.8158 Iter 449, loss: -725.6671 Iter 450, loss: -728.4734 Iter 451, loss: -731.3931 Iter 452, loss: -734.2934 Iter 453, loss: -737.0539 Iter 454, loss: -739.8947 Iter 455, loss: -742.8338 Iter 456, loss: -745.6566 Iter 457, loss: -748.4380 Iter 458, loss: -751.3253 Iter 459, loss: -754.2007 Iter 460, loss: -757.0026

Iter 461, loss: -759.8360 Iter 462, loss: -762.6934 Iter 463, loss: -765.5159 Iter 464, loss: -768.3442 Iter 465, loss: -771.1942 Iter 466, loss: -774.0161 Iter 467, loss: -776.8235 Iter 468, loss: -779.6633 Iter 469, loss: -782.5107 Iter 470, loss: -785.3283 Iter 471, loss: -788.1385 Iter 472, loss: -790.9679 Iter 473, loss: -793.7994 Iter 474, loss: -796.6173 Iter 475, loss: -799.4358 Iter 476, loss: -802.2635 Iter 477, loss: -805.0873 Iter 478, loss: -807.8990 Iter 479, loss: -810.7078 Iter 480, loss: -813.5228 Iter 481, loss: -816.3396 Iter 482, loss: -819.1496 Iter 483, loss: -821.9536 Iter 484, loss: -824.7565 Iter 485, loss: -827.5608 Iter 486, loss: -830.3618

Iter 487, loss: -833.1549 Iter 488, loss: -835.9385 Iter 489, loss: -838.7115 Iter 490, loss: -841.4657 Iter 491, loss: -844.1868 Iter 492, loss: -846.8416 Iter 493, loss: -849.4214 Iter 494, loss: -851.9346 Iter 495, loss: -854.6054 Iter 496, loss: -857.5202 Iter 497, loss: -860.5240 Iter 498, loss: -863.2225 Iter 499, loss: -865.6365 Iter 500, loss: -868.1282 Iter 501, loss: -870.8827 Iter 502, loss: -873.7051 Iter 503, loss: -876.2942 Iter 504, loss: -878.7282 Iter 505, loss: -881.2866 Iter 506, loss: -884.0038 Iter 507, loss: -886.6721 Iter 508, loss: -889.1764 Iter 509, loss: -891.6511 Iter 510, loss: -894.2421 Iter 511, loss: -896.8917 Iter 512, loss: -899.4667

Iter 513, loss: -901.9542 Iter 514, loss: -904.4536 Iter 515, loss: -907.0261 Iter 516, loss: -909.6190 Iter 517, loss: -912.1621 Iter 518, loss: -914.6556 Iter 519, loss: -917.1533 Iter 520, loss: -919.6907 Iter 521, loss: -922.2477 Iter 522, loss: -924.7869 Iter 523, loss: -927.2935 Iter 524, loss: -929.7855 Iter 525, loss: -932.2878 Iter 526, loss: -934.8078 Iter 527, loss: -937.3365 Iter 528, loss: -939.8578 Iter 529, loss: -942.3646 Iter 530, loss: -944.8597 Iter 531, loss: -947.3508 Iter 532, loss: -949.8453 Iter 533, loss: -952.3446 Iter 534, loss: -954.8483 Iter 535, loss: -957.3527 Iter 536, loss: -959.8552 Iter 537, loss: -962.3537 Iter 538, loss: -964.8478

Iter 539, loss: -967.3373 Iter 540, loss: -969.8223 Iter 541, loss: -972.3016 Iter 542, loss: -974.7719 Iter 543, loss: -977.2246 Iter 544, loss: -979.6384 Iter 545, loss: -981.9717 Iter 546, loss: -984.1694 Iter 547, loss: -986.3162 Iter 548, loss: -988.7554 Iter 549, loss: -991.5734 Iter 550, loss: -994.1413 Iter 551, loss: -996.2031 Iter 552, loss: -998.3425 Iter 553, loss: -1000.9259 Iter 554, loss: -1003.4068 Iter 555, loss: -1005.4825 Iter 556, loss: -1007.7172 Iter 557, loss: -1010.2598 Iter 558, loss: -1012.5682 Iter 559, loss: -1014.6837 Iter 560, loss: -1017.0267 Iter 561, loss: -1019.4419 Iter 562, loss: -1021.6270 Iter 563, loss: -1023.8052

Iter 564, loss: -1026.1648 Iter 565, loss: -1028.4509 Iter 566, loss: -1030.5997 Iter 567, loss: -1032.8464 Iter 568, loss: -1035.1584 Iter 569, loss: -1037.3605 Iter 570, loss: -1039.5227 Iter 571, loss: -1041.7581 Iter 572, loss: -1043.9795 Iter 573, loss: -1046.1007 Iter 574, loss: -1048.2045 Iter 575, loss: -1050.3597 Iter 576, loss: -1052.4878 Iter 577, loss: -1054.6168 Iter 578, loss: -1056.8713 Iter 579, loss: -1059.2063 Iter 580, loss: -1061.4543 Iter 581, loss: -1063.6030 Iter 582, loss: -1065.6982 Iter 583, loss: -1067.7516 Iter 584, loss: -1069.7948 Iter 585, loss: -1071.9045 Iter 586, loss: -1074.0956 Iter 587, loss: -1076.2740 Iter 588, loss: -1078.3936

Iter 589, loss: -1080.4835 Iter 590, loss: -1082.5472 Iter 591, loss: -1084.5833 Iter 592, loss: -1086.6300 Iter 593, loss: -1088.7178 Iter 594, loss: -1090.8267 Iter 595, loss: -1092.9258 Iter 596, loss: -1095.0076 Iter 597, loss: -1097.0801 Iter 598, loss: -1099.1362 Iter 599, loss: -1101.1686 Iter 600, loss: -1103.1925 Iter 601, loss: -1105.2222 Iter 602, loss: -1107.2546 Iter 603, loss: -1109.2836 Iter 604, loss: -1111.3105 Iter 605, loss: -1113.3396 Iter 606, loss: -1115.3721 Iter 607, loss: -1117.4001 Iter 608, loss: -1119.4213 Iter 609, loss: -1121.4374 Iter 610, loss: -1123.4521 Iter 611, loss: -1125.4623 Iter 612, loss: -1127.4648 Iter 613, loss: -1129.4585 Iter 614, loss: -1131.4457

Iter 615, loss: -1133.4255 Iter 616, loss: -1135.3960 Iter 617, loss: -1137.3506 Iter 618, loss: -1139.2836 Iter 619, loss: -1141.1881 Iter 620, loss: -1143.0603 Iter 621, loss: -1144.8972 Iter 622, loss: -1146.7253 Iter 623, loss: -1148.5906 Iter 624, loss: -1150.5715 Iter 625, loss: -1152.6666 Iter 626, loss: -1154.7900 Iter 627, loss: -1156.8048 Iter 628, loss: -1158.6639 Iter 629, loss: -1160.4279 Iter 630, loss: -1162.1996 Iter 631, loss: -1164.0623 Iter 632, loss: -1166.0195 Iter 633, loss: -1168.0066 Iter 634, loss: -1169.9358 Iter 635, loss: -1171.7760 Iter 636, loss: -1173.5613 Iter 637, loss: -1175.3496 Iter 638, loss: -1177.1837 Iter 639, loss: -1179.0658 Iter 640, loss: -1180.9668

Iter 641, loss: -1182.8463 Iter 642, loss: -1184.6846 Iter 643, loss: -1186.4877 Iter 644, loss: -1188.2764 Iter 645, loss: -1190.0714 Iter 646, loss: -1191.8834 Iter 647, loss: -1193.7133 Iter 648, loss: -1195.5508 Iter 649, loss: -1197.3859 Iter 650, loss: -1199.2094 Iter 651, loss: -1201.0181 Iter 652, loss: -1202.8136 Iter 653, loss: -1204.5996 Iter 654, loss: -1206.3798 Iter 655, loss: -1208.1561 Iter 656, loss: -1209.9299 Iter 657, loss: -1211.7008 Iter 658, loss: -1213.4688 Iter 659, loss: -1215.2319 Iter 660, loss: -1216.9904 Iter 661, loss: -1218.7418 Iter 662, loss: -1220.4873 Iter 663, loss: -1222.2244 Iter 664, loss: -1223.9561 Iter 665, loss: -1225.6816 Iter 666, loss: -1227.4078

Iter 667, loss: -1229.1348 Iter 668, loss: -1230.8716 Iter 669, loss: -1232.6161 Iter 670, loss: -1234.3710 Iter 671, loss: -1236.1263 Iter 672, loss: -1237.8766 Iter 673, loss: -1239.6101 Iter 674, loss: -1241.3213 Iter 675, loss: -1243.0052 Iter 676, loss: -1244.6597 Iter 677, loss: -1246.2816 Iter 678, loss: -1247.8696 Iter 679, loss: -1249.4310 Iter 680, loss: -1250.9948 Iter 681, loss: -1252.6107 Iter 682, loss: -1254.3292 Iter 683, loss: -1256.1324 Iter 684, loss: -1257.9323 Iter 685, loss: -1259.6361 Iter 686, loss: -1261.2280 Iter 687, loss: -1262.7694 Iter 688, loss: -1264.3458 Iter 689, loss: -1266.0016 Iter 690, loss: -1267.7037 Iter 691, loss: -1269.3749

Iter 692, loss: -1270.9707 Iter 693, loss: -1272.5109 Iter 694, loss: -1274.0526 Iter 695, loss: -1275.6433 Iter 696, loss: -1277.2833 Iter 697, loss: -1278.9287 Iter 698, loss: -1280.5336 Iter 699, loss: -1282.0913 Iter 700, loss: -1283.6311 Iter 701, loss: -1285.1870 Iter 702, loss: -1286.7721 Iter 703, loss: -1288.3726 Iter 704, loss: -1289.9623 Iter 705, loss: -1291.5236 Iter 706, loss: -1293.0563 Iter 707, loss: -1294.5703 Iter 708, loss: -1296.0757 Iter 709, loss: -1297.5723 Iter 710, loss: -1299.0502 Iter 711, loss: -1300.4885 Iter 712, loss: -1301.8833 Iter 713, loss: -1303.2554 Iter 714, loss: -1304.6912 Iter 715, loss: -1306.2898 Iter 716, loss: -1308.0707 Iter 717, loss: -1309.8831

Iter 718, loss: -1311.5248 Iter 719, loss: -1312.9395 Iter 720, loss: -1314.2437 Iter 721, loss: -1315.6068 Iter 722, loss: -1317.1259 Iter 723, loss: -1318.7583 Iter 724, loss: -1320.3695 Iter 725, loss: -1321.8630 Iter 726, loss: -1323.2579 Iter 727, loss: -1324.6443 Iter 728, loss: -1326.0911 Iter 729, loss: -1327.5980 Iter 730, loss: -1329.1132 Iter 731, loss: -1330.5875 Iter 732, loss: -1332.0168 Iter 733, loss: -1333.4320 Iter 734, loss: -1334.8649 Iter 735, loss: -1336.3220 Iter 736, loss: -1337.7843 Iter 737, loss: -1339.2284 Iter 738, loss: -1340.6461 Iter 739, loss: -1342.0471 Iter 740, loss: -1343.4491 Iter 741, loss: -1344.8651 Iter 742, loss: -1346.2965 Iter 743, loss: -1347.7330

Iter 744, loss: -1349.1620 Iter 745, loss: -1350.5750 Iter 746, loss: -1351.9725 Iter 747, loss: -1353.3607 Iter 748, loss: -1354.7461 Iter 749, loss: -1356.1334 Iter 750, loss: -1357.5244 Iter 751, loss: -1358.9187 Iter 752, loss: -1360.3125 Iter 753, loss: -1361.7029 Iter 754, loss: -1363.0868 Iter 755, loss: -1364.4618 Iter 756, loss: -1365.8252 Iter 757, loss: -1367.1731 Iter 758, loss: -1368.4995 Iter 759, loss: -1369.7938 Iter 760, loss: -1371.0430 Iter 761, loss: -1372.2306 Iter 762, loss: -1373.3635 Iter 763, loss: -1374.4858 Iter 764, loss: -1375.7313 Iter 765, loss: -1377.2146 Iter 766, loss: -1378.9100 Iter 767, loss: -1380.5641 Iter 768, loss: -1381.9559

Iter 769, loss: -1383.1044 Iter 770, loss: -1384.2153 Iter 771, loss: -1385.4858 Iter 772, loss: -1386.9238 Iter 773, loss: -1388.3632 Iter 774, loss: -1389.6580 Iter 775, loss: -1390.8379 Iter 776, loss: -1392.0469 Iter 777, loss: -1393.3768 Iter 778, loss: -1394.7676 Iter 779, loss: -1396.0896 Iter 780, loss: -1397.2997 Iter 781, loss: -1398.4716 Iter 782, loss: -1399.7041 Iter 783, loss: -1401.0164 Iter 784, loss: -1402.3422 Iter 785, loss: -1403.6101 Iter 786, loss: -1404.8193 Iter 787, loss: -1406.0250 Iter 788, loss: -1407.2715 Iter 789, loss: -1408.5508 Iter 790, loss: -1409.8220 Iter 791, loss: -1411.0574 Iter 792, loss: -1412.2638 Iter 793, loss: -1413.4705 Iter 794, loss: -1414.6963

Iter 795, loss: -1415.9403 Iter 796, loss: -1417.1833 Iter 797, loss: -1418.4098 Iter 798, loss: -1419.6163 Iter 799, loss: -1420.8118 Iter 800, loss: -1422.0066 Iter 801, loss: -1423.2041 Iter 802, loss: -1424.3976 Iter 803, loss: -1425.5746 Iter 804, loss: -1426.7218 Iter 805, loss: -1427.8273 Iter 806, loss: -1428.8918 Iter 807, loss: -1429.9255 Iter 808, loss: -1430.9750 Iter 809, loss: -1432.1062 Iter 810, loss: -1433.3989 Iter 811, loss: -1434.8374 Iter 812, loss: -1436.3063 Iter 813, loss: -1437.6500 Iter 814, loss: -1438.8112 Iter 815, loss: -1439.8541 Iter 816, loss: -1440.8990 Iter 817, loss: -1442.0469 Iter 818, loss: -1443.3158 Iter 819, loss: -1444.6416 Iter 820, loss: -1445.9211

Iter 821, loss: -1447.0989 Iter 822, loss: -1448.1952 Iter 823, loss: -1449.2769 Iter 824, loss: -1450.4041 Iter 825, loss: -1451.5953 Iter 826, loss: -1452.8253 Iter 827, loss: -1454.0465 Iter 828, loss: -1455.2256 Iter 829, loss: -1456.3605 Iter 830, loss: -1457.4735 Iter 831, loss: -1458.5933 Iter 832, loss: -1459.7366 Iter 833, loss: -1460.9038 Iter 834, loss: -1462.0825 Iter 835, loss: -1463.2557 Iter 836, loss: -1464.4121 Iter 837, loss: -1465.5493 Iter 838, loss: -1466.6736 Iter 839, loss: -1467.7927 Iter 840, loss: -1468.9146 Iter 841, loss: -1470.0433 Iter 842, loss: -1471.1796 Iter 843, loss: -1472.3215 Iter 844, loss: -1473.4661 Iter 845, loss: -1474.6099 Iter 846, loss: -1475.7506

Iter 847, loss: -1476.8868 Iter 848, loss: -1478.0168 Iter 849, loss: -1479.1376 Iter 850, loss: -1480.2430 Iter 851, loss: -1481.3209 Iter 852, loss: -1482.3516 Iter 853, loss: -1483.3075 Iter 854, loss: -1484.1731 Iter 855, loss: -1484.9980 Iter 856, loss: -1485.9451 Iter 857, loss: -1487.2031 Iter 858, loss: -1488.6915 Iter 859, loss: -1490.0686 Iter 860, loss: -1491.1125 Iter 861, loss: -1491.9501 Iter 862, loss: -1492.8811 Iter 863, loss: -1494.0679 Iter 864, loss: -1495.3617 Iter 865, loss: -1496.4805 Iter 866, loss: -1497.3683 Iter 867, loss: -1498.2461 Iter 868, loss: -1499.3169 Iter 869, loss: -1500.5140 Iter 870, loss: -1501.6099 Iter 871, loss: -1502.5405 Iter 872, loss: -1503.4691

Iter 873, loss: -1504.5277 Iter 874, loss: -1505.6548 Iter 875, loss: -1506.7106 Iter 876, loss: -1507.6720 Iter 877, loss: -1508.6316 Iter 878, loss: -1509.6552 Iter 879, loss: -1510.7156 Iter 880, loss: -1511.7483 Iter 881, loss: -1512.7325 Iter 882, loss: -1513.7008 Iter 883, loss: -1514.6898 Iter 884, loss: -1515.7070 Iter 885, loss: -1516.7262 Iter 886, loss: -1517.7213 Iter 887, loss: -1518.6901 Iter 888, loss: -1519.6525 Iter 889, loss: -1520.6271 Iter 890, loss: -1521.6085 Iter 891, loss: -1522.5718 Iter 892, loss: -1523.4933 Iter 893, loss: -1524.3640 Iter 894, loss: -1525.1801 Iter 895, loss: -1525.9376 Iter 896, loss: -1526.6296 Iter 897, loss: -1527.3164 Iter 898, loss: -1528.1212

Iter 899, loss: -1529.2394 Iter 900, loss: -1530.6628 Iter 901, loss: -1532.1213 Iter 902, loss: -1533.2657 Iter 903, loss: -1534.0319 Iter 904, loss: -1534.6592 Iter 905, loss: -1535.4536 Iter 906, loss: -1536.5481 Iter 907, loss: -1537.7806 Iter 908, loss: -1538.8776 Iter 909, loss: -1539.7313 Iter 910, loss: -1540.4736 Iter 911, loss: -1541.3102 Iter 912, loss: -1542.3206 Iter 913, loss: -1543.4083 Iter 914, loss: -1544.4150 Iter 915, loss: -1545.2853 Iter 916, loss: -1546.0967 Iter 917, loss: -1546.9618 Iter 918, loss: -1547.9221 Iter 919, loss: -1548.9233 Iter 920, loss: -1549.8842 Iter 921, loss: -1550.7723 Iter 922, loss: -1551.6222 Iter 923, loss: -1552.4908 Iter 924, loss: -1553.4061

Iter 925, loss: -1554.3521 Iter 926, loss: -1555.2904 Iter 927, loss: -1556.1951 Iter 928, loss: -1557.0696 Iter 929, loss: -1557.9374 Iter 930, loss: -1558.8198 Iter 931, loss: -1559.7230 Iter 932, loss: -1560.6368 Iter 933, loss: -1561.5452 Iter 934, loss: -1562.4381 Iter 935, loss: -1563.3165 Iter 936, loss: -1564.1886 Iter 937, loss: -1565.0631 Iter 938, loss: -1565.9449 Iter 939, loss: -1566.8333 Iter 940, loss: -1567.7234 Iter 941, loss: -1568.6106 Iter 942, loss: -1569.4919 Iter 943, loss: -1570.3667 Iter 944, loss: -1571.2367 Iter 945, loss: -1572.1047 Iter 946, loss: -1572.9725 Iter 947, loss: -1573.8413 Iter 948, loss: -1574.7111 Iter 949, loss: -1575.5815

Iter 950, loss: -1576.4514 Iter 951, loss: -1577.3206 Iter 952, loss: -1578.1869 Iter 953, loss: -1579.0513 Iter 954, loss: -1579.9133 Iter 955, loss: -1580.7723 Iter 956, loss: -1581.6279 Iter 957, loss: -1582.4801 Iter 958, loss: -1583.3258 Iter 959, loss: -1584.1624 Iter 960, loss: -1584.9833 Iter 961, loss: -1585.7794 Iter 962, loss: -1586.5341 Iter 963, loss: -1587.2338 Iter 964, loss: -1587.8776 Iter 965, loss: -1588.5223 Iter 966, loss: -1589.2742 Iter 967, loss: -1590.2296 Iter 968, loss: -1591.3195 Iter 969, loss: -1592.3779 Iter 970, loss: -1593.2877 Iter 971, loss: -1594.0546 Iter 972, loss: -1594.7512 Iter 973, loss: -1595.4644 Iter 974, loss: -1596.2758 Iter 975, loss: -1597.2021

Iter 976, loss: -1598.1466 Iter 977, loss: -1598.9785 Iter 978, loss: -1599.6815 Iter 979, loss: -1600.3694 Iter 980, loss: -1601.1519 Iter 981, loss: -1602.0186 Iter 982, loss: -1602.8639 Iter 983, loss: -1603.6154 Iter 984, loss: -1604.2891 Iter 985, loss: -1604.9398 Iter 986, loss: -1605.5845 Iter 987, loss: -1606.2009 Iter 988, loss: -1606.7559 Iter 989, loss: -1607.2969 Iter 990, loss: -1607.9141 Iter 991, loss: -1608.7660 Iter 992, loss: -1609.8618 Iter 993, loss: -1611.0719 Iter 994, loss: -1612.1476 Iter 995, loss: -1612.9281 Iter 996, loss: -1613.4553 Iter 997, loss: -1613.9492 Iter 998, loss: -1614.6454 Iter 999, loss: -1615.5844

Remember that the score function parametrizes the gradient of the log, so it is a vector valued function. So we can use a little helper function to visualize the score function we have just fitted as a vector field:

def plot_vector_field(model, ax=None, x_range=(-2, 2, 10), y_range=(-2, 2, 10), **kwargs):
    """Plot vector field of grad(log(p)), aka score function"""
    if ax is None:
        ax = plt.gca()
    
    xx, yy = jnp.meshgrid(
        jnp.linspace(*x_range),
        jnp.linspace(*y_range)
    )
    grid = jnp.stack([xx.ravel(), yy.ravel()], axis=-1)

    scores = jax.vmap(model)(grid)

    scores_norm = jnp.linalg.norm(scores, axis=-1, ord=2, keepdims=True)
    scores_log1p = scores / scores_norm * jnp.log(1 + scores_norm)
    ax.quiver(xx, yy, scores_log1p[:, 0], scores_log1p[:, 1], **kwargs)
    return ax

ax = plt.subplot()

ax.scatter(x[:, 0], x[:, 1], s=10, alpha=0.1)

ax.set_aspect('equal')
scale = 2.0
ax.set_xlim(-scale, scale)
ax.set_ylim(-scale, scale)
ax.set_xlabel("x")
ax.set_ylabel("y")
plot_vector_field(model, ax=ax, x_range=(-2, 2, 20), y_range=(-2, 2, 20), color="red");

png

If you look closely you can see that the gradient always points towards the direction of the highest probability density at any point in the data space. Which is exactly what we expect from intuition.

Now that we have the score function of the distribution we can actually go back to our Langevin sampling method and replace the analytical score function with our fitted model:

Args = namedtuple("Args", ["key", "idx", "x", "alpha_0", "p_0"])

def sample(score, args, _):
    key, subkey = random.split(args.key)
    alpha = args.alpha_0 * (args.p_0 ** args.idx) ** 2
    dx = random.normal(subkey, args.x.shape)
    x = args.x + 0.5 * alpha * jax.vmap(score)(args.x) + jnp.sqrt(alpha) * dx
    return Args(key, args.idx + 1, x, args.alpha_0, args.p_0), x


n_samples = 10_000
n_iter = 2000

key = random.PRNGKey(42)
key, subkey = random.split(key)

init = Args(
    key=key,
    idx=0,
    x=random.normal(subkey, (n_samples, 2)),
    alpha_0=0.0001,
    p_0=1,
)

result, sample_trace = jax.lax.scan(partial(sample, model), init, length=n_iter)

This is what the samples look like:

ax = plt.subplot()

ax.scatter(sample_trace[-1, :, 0], sample_trace[-1, :, 1], s=1)
ax.set_aspect('equal')
scale = 2.0
ax.set_xlim(-scale, scale)
ax.set_ylim(-scale, scale)
ax.set_xlabel("x")
ax.set_ylabel("y");

png

(Break)

Denoising Diffusion Probabilistic Models (DDPM)

The previous result just using score matching was not bad, but also not great either. The sampling process got “stuck” in regions of lower probability and missed details of the distribution. How can we improve upon this?

The answer that will finally bring us to the DDPM “diffusion model” has multiple aspects. The most important ideas is, we can make the model of the score function dependent (conditioned) on time. But how does the training of such a model look like? We can create “diffused” versions of our training data as we did above, which change with time and move (or “morph”) from a simple Gaussian to the much more complex target distribution.

Practically we can make another small improvement to the diffusion process: in order to get the “diffused” example data a given $t$, we had to compute all the previous steps (check the simple example above). Now, with a bit of math, because all the transitions between the different stages are Gaussian, one can derive a closed form expression to how much the data is diffused at given time $t$ (check the DDPM paper).

The solution is given by:

$q(x_t x_0) = \sqrt{\bar{\alpha_t}}x_0 + (1 - \bar{\alpha_t}) \cdot \epsilon$

And

$\bar{\alpha} = (1 - \beta)^t, \textrm{with~} \alpha = 1 - \beta$

Now let’s first define the range of $\beta$ values and compute $\alpha$ and $\bar{\alpha}$ from it:

n_t_steps = 25 # number of time steps
betas = jnp.geomspace(1e-4, 0.9999, n_t_steps)
alphas = 1 - betas.reshape(1, -1, 1)
alpha_bars = jnp.cumprod(alphas, axis=1)

This we can use to diffuse the data:

noise = jax.random.normal(jax.random.PRNGKey(1234), shape=(x.shape[0], n_t_steps, 2))
x_diffused_example = jnp.sqrt(alpha_bars) * x[:, jnp.newaxis, :] + jnp.sqrt(1 - alpha_bars) * noise

Now we can visualize the diffusion process with time:

fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(8, 8))

for idx, ax in enumerate(axes.flat):
    ax.scatter(x_diffused_example[:, idx, 0], x_diffused_example[:, idx, 1], s=1)
    ax.set_title(f"Step {idx}", fontdict={"size": 10}, pad=3)
    ax.set_xticks([])
    ax.set_yticks([])

png

We can quickly verify that just as in the other example above the marginal distributions correspond to a unit Gaussian:

ax = plt.subplot()

ax.hist(x_diffused_example[:, -1, 0], density=True, bins=50, label="X Marginal", histtype="step")
ax.hist(x_diffused_example[:, -1, 1], density=True, bins=50, label="Y Marginal", histtype="step")


x_plot = jnp.linspace(-5, 5, 100)
ax.plot(x_plot, jax.scipy.stats.norm.pdf(x_plot), label="Unit Gaussian")
ax.set_xlabel("x")
ax.set_ylabel("PDF")
ax.legend();

png

Now we modify our model to be dependent on the time coordinate. A common way to do this is via a so called “positional embedding”. For this we consider the time variable to be an integer value, and the embedding consists of a matrix of the size $N_t \times N_{out}$, where $N_t$ is the number of time steps to consider and $N_{out}$ is the number of dimensions in the intermediate layer of the MLP.

The given a time idx t we can just lookup the vector in the embedding and directly mutliply it to the intermediate activations:

class TimeConditionedLinear(eqx.Module):
    """Simple MLP"""
    linear: eqx.nn.Linear
    embedding: eqx.nn.Embedding

    def __init__(self, in_dim, out_dim, t_dim, key):
        keys = jax.random.split(key, 2)
        self.embedding = eqx.nn.Embedding(t_dim, out_dim, key=keys[0])
        self.linear = eqx.nn.Linear(in_dim, out_dim, key=keys[1])

    def __call__(self, x, t):
        out = self.linear(x)
        value = jax.vmap(self.embedding)(t)[0]
        result = out * value
        return result


class TimeConditionedMLP(eqx.Module):
    """Simple MLP"""
    fc1: TimeConditionedLinear
    fc2: TimeConditionedLinear
    fc3: eqx.nn.Linear

    def __init__(self, in_dim, out_dim, t_dim, key, hidden_dim=128):
        keys = jax.random.split(key, 3)
        self.fc1 = TimeConditionedLinear(in_dim, hidden_dim, t_dim, key=keys[0])
        self.fc2 = TimeConditionedLinear(hidden_dim, hidden_dim, t_dim, key=keys[1])
        self.fc3 = eqx.nn.Linear(hidden_dim, out_dim, key=keys[2])

    def __call__(self, x, t):
        x = jax.nn.gelu(self.fc1(x, t))
        x = jax.nn.gelu(self.fc2(x, t))
        x = self.fc3(x)
        return x

In the training we will apply one more additional modification. We define the model and loss function such that the model predicts the noise in each step. There is a nice but somewhat lenghty derivation in the Deepia video above and in the DDPM paper.

But just intuitively: the model is trained to predict the noise in each step. In the sampling process we can then just subtract the noise from the current iteration and get a less noisy version of the data. If we do this iteratively we arrive at a final denoised sample of the data distribution, matching the process illustrated in the figure at the top.

Ok let’s implement this in code:

n_t_steps = 100 # number of time steps
betas = jnp.geomspace(1e-4, 0.9999, n_t_steps)
alphas = 1.0 - betas
alpha_bars = jnp.cumprod(alphas)

model_cond = TimeConditionedMLP(2, 2, t_dim=n_t_steps, key=jax.random.key(23))

optimizer = optax.adam(learning_rate=1e-3)
opt_state = optimizer.init(model_cond)

n_iter = 20_000


def mse_loss(model, x, t, noise):
    """Mean squared losss"""
    pred = jax.vmap(model)(x, t)
    return ((pred - noise) ** 2).mean()


@eqx.filter_jit
def make_step(model, opt_state, x, t, noise):
    """Make a single update step"""
    loss_value, grads = eqx.filter_value_and_grad(mse_loss)(model, x, t, noise)
    updates, opt_state = optimizer.update(
        grads, opt_state, eqx.filter(model, eqx.is_array)
    )
    model = eqx.apply_updates(model, updates)
    return model, opt_state, loss_value


key = jax.random.PRNGKey(1234)

for idx in range(n_iter):
    key, subkey = random.split(key)
    noise = jax.random.normal(subkey, shape=x.shape)

    key, subkey = random.split(key)
    t = jax.random.randint(subkey, minval=0, maxval=n_t_steps, shape=(x.shape[0], 1))

    x_diffused = jnp.sqrt(alpha_bars[t]) * x + jnp.sqrt(1 - alpha_bars[t]) * noise

    model_cond, opt_state, train_loss = make_step(model_cond, opt_state, x_diffused, t, noise)
    print(f"Iter {idx}, loss: {train_loss:.4f}", end="\r")

Iter 0, loss: 0.9840 Iter 1, loss: 0.9978 Iter 2, loss: 0.9667 Iter 3, loss: 0.9926 Iter 4, loss: 0.9654 Iter 5, loss: 0.9440 Iter 6, loss: 0.9776 Iter 7, loss: 0.9805 Iter 8, loss: 0.9519 Iter 9, loss: 0.9188 Iter 10, loss: 1.0504 Iter 11, loss: 0.9044 Iter 12, loss: 0.9489 Iter 13, loss: 0.9364 Iter 14, loss: 0.9026 Iter 15, loss: 0.9090 Iter 16, loss: 0.9580 Iter 17, loss: 0.9814 Iter 18, loss: 0.8847 Iter 19, loss: 0.8942 Iter 20, loss: 0.8895 Iter 21, loss: 0.8822 Iter 22, loss: 0.8895 Iter 23, loss: 0.8406 Iter 24, loss: 0.8471 Iter 25, loss: 0.8708 Iter 26, loss: 0.9356 Iter 27, loss: 0.8707 Iter 28, loss: 0.8433 Iter 29, loss: 0.8098 Iter 30, loss: 0.8481 Iter 31, loss: 0.8342 Iter 32, loss: 0.8431 Iter 33, loss: 0.8070 Iter 34, loss: 0.8239 Iter 35, loss: 0.7923 Iter 36, loss: 0.8325 Iter 37, loss: 0.8015 Iter 38, loss: 0.7957 Iter 39, loss: 0.8296 Iter 40, loss: 0.7984

Iter 41, loss: 0.7781 Iter 42, loss: 0.7673 Iter 43, loss: 0.8154 Iter 44, loss: 0.7794 Iter 45, loss: 0.7998 Iter 46, loss: 0.7494 Iter 47, loss: 0.7387 Iter 48, loss: 0.7364 Iter 49, loss: 0.7280 Iter 50, loss: 0.7886 Iter 51, loss: 0.7448 Iter 52, loss: 0.7545 Iter 53, loss: 0.7375 Iter 54, loss: 0.7353 Iter 55, loss: 0.7798 Iter 56, loss: 0.8203 Iter 57, loss: 0.6955 Iter 58, loss: 0.7414 Iter 59, loss: 0.6853 Iter 60, loss: 0.7088 Iter 61, loss: 0.7057 Iter 62, loss: 0.7141 Iter 63, loss: 0.6953 Iter 64, loss: 0.7040 Iter 65, loss: 0.6626 Iter 66, loss: 0.7030 Iter 67, loss: 0.7011 Iter 68, loss: 0.7074 Iter 69, loss: 0.6634 Iter 70, loss: 0.7049 Iter 71, loss: 0.7018 Iter 72, loss: 0.6776 Iter 73, loss: 0.7339 Iter 74, loss: 0.6815 Iter 75, loss: 0.6861 Iter 76, loss: 0.6817 Iter 77, loss: 0.6882 Iter 78, loss: 0.6769 Iter 79, loss: 0.6488 Iter 80, loss: 0.7011 Iter 81, loss: 0.6507

Iter 82, loss: 0.6816 Iter 83, loss: 0.6478 Iter 84, loss: 0.6817 Iter 85, loss: 0.6831 Iter 86, loss: 0.6878 Iter 87, loss: 0.6873 Iter 88, loss: 0.6736 Iter 89, loss: 0.6708 Iter 90, loss: 0.6649 Iter 91, loss: 0.6776 Iter 92, loss: 0.6793 Iter 93, loss: 0.6946 Iter 94, loss: 0.6601 Iter 95, loss: 0.6835 Iter 96, loss: 0.6482 Iter 97, loss: 0.6248 Iter 98, loss: 0.6646 Iter 99, loss: 0.6358 Iter 100, loss: 0.6348 Iter 101, loss: 0.6909 Iter 102, loss: 0.6511 Iter 103, loss: 0.6395 Iter 104, loss: 0.6439 Iter 105, loss: 0.6458 Iter 106, loss: 0.7220 Iter 107, loss: 0.6693 Iter 108, loss: 0.6868 Iter 109, loss: 0.6775 Iter 110, loss: 0.6945 Iter 111, loss: 0.6666 Iter 112, loss: 0.6386 Iter 113, loss: 0.6450 Iter 114, loss: 0.6794 Iter 115, loss: 0.6498 Iter 116, loss: 0.6191 Iter 117, loss: 0.6644 Iter 118, loss: 0.6718 Iter 119, loss: 0.6332 Iter 120, loss: 0.6370 Iter 121, loss: 0.6627 Iter 122, loss: 0.6774

Iter 123, loss: 0.6581 Iter 124, loss: 0.6587 Iter 125, loss: 0.6758 Iter 126, loss: 0.6912 Iter 127, loss: 0.6662 Iter 128, loss: 0.6239 Iter 129, loss: 0.6085 Iter 130, loss: 0.6076 Iter 131, loss: 0.6420 Iter 132, loss: 0.6111 Iter 133, loss: 0.6593 Iter 134, loss: 0.6156 Iter 135, loss: 0.6735 Iter 136, loss: 0.6645 Iter 137, loss: 0.6629 Iter 138, loss: 0.6199 Iter 139, loss: 0.6649 Iter 140, loss: 0.6999 Iter 141, loss: 0.6520 Iter 142, loss: 0.6611 Iter 143, loss: 0.6456 Iter 144, loss: 0.6776 Iter 145, loss: 0.6486 Iter 146, loss: 0.6577 Iter 147, loss: 0.6527 Iter 148, loss: 0.6524 Iter 149, loss: 0.6540 Iter 150, loss: 0.6561 Iter 151, loss: 0.6793 Iter 152, loss: 0.6447 Iter 153, loss: 0.6181 Iter 154, loss: 0.6658 Iter 155, loss: 0.6534 Iter 156, loss: 0.6422 Iter 157, loss: 0.6766 Iter 158, loss: 0.6452 Iter 159, loss: 0.6989 Iter 160, loss: 0.6520 Iter 161, loss: 0.6371 Iter 162, loss: 0.6440 Iter 163, loss: 0.6414

Iter 164, loss: 0.6796 Iter 165, loss: 0.6573 Iter 166, loss: 0.6752 Iter 167, loss: 0.6531 Iter 168, loss: 0.6171 Iter 169, loss: 0.6415 Iter 170, loss: 0.6796 Iter 171, loss: 0.6314 Iter 172, loss: 0.6558 Iter 173, loss: 0.6500 Iter 174, loss: 0.6725 Iter 175, loss: 0.6095 Iter 176, loss: 0.6550 Iter 177, loss: 0.6838 Iter 178, loss: 0.6450 Iter 179, loss: 0.6579 Iter 180, loss: 0.6577 Iter 181, loss: 0.6345 Iter 182, loss: 0.6914 Iter 183, loss: 0.6508 Iter 184, loss: 0.6407 Iter 185, loss: 0.6606 Iter 186, loss: 0.6642 Iter 187, loss: 0.6567 Iter 188, loss: 0.6634 Iter 189, loss: 0.6246 Iter 190, loss: 0.6696 Iter 191, loss: 0.6369 Iter 192, loss: 0.6486 Iter 193, loss: 0.6475 Iter 194, loss: 0.6565 Iter 195, loss: 0.6166 Iter 196, loss: 0.6095 Iter 197, loss: 0.6377 Iter 198, loss: 0.6178 Iter 199, loss: 0.6251 Iter 200, loss: 0.6452 Iter 201, loss: 0.6476 Iter 202, loss: 0.6580 Iter 203, loss: 0.6422 Iter 204, loss: 0.6498

Iter 205, loss: 0.6230 Iter 206, loss: 0.6427 Iter 207, loss: 0.6173 Iter 208, loss: 0.6785 Iter 209, loss: 0.6581 Iter 210, loss: 0.6612 Iter 211, loss: 0.6223 Iter 212, loss: 0.6505 Iter 213, loss: 0.6912 Iter 214, loss: 0.6331 Iter 215, loss: 0.6287 Iter 216, loss: 0.6288 Iter 217, loss: 0.6394 Iter 218, loss: 0.6914 Iter 219, loss: 0.6744 Iter 220, loss: 0.6740 Iter 221, loss: 0.6369 Iter 222, loss: 0.6734 Iter 223, loss: 0.5976 Iter 224, loss: 0.6312 Iter 225, loss: 0.6925 Iter 226, loss: 0.6667 Iter 227, loss: 0.6861 Iter 228, loss: 0.6754 Iter 229, loss: 0.6702 Iter 230, loss: 0.6273 Iter 231, loss: 0.6556 Iter 232, loss: 0.6395 Iter 233, loss: 0.6280 Iter 234, loss: 0.6259 Iter 235, loss: 0.6569 Iter 236, loss: 0.6472 Iter 237, loss: 0.6169 Iter 238, loss: 0.6704 Iter 239, loss: 0.6400 Iter 240, loss: 0.6530 Iter 241, loss: 0.6662 Iter 242, loss: 0.6623 Iter 243, loss: 0.6497 Iter 244, loss: 0.6716 Iter 245, loss: 0.6508 Iter 246, loss: 0.6350

Iter 247, loss: 0.5989 Iter 248, loss: 0.6138 Iter 249, loss: 0.6562 Iter 250, loss: 0.6676 Iter 251, loss: 0.6434 Iter 252, loss: 0.6560 Iter 253, loss: 0.6756 Iter 254, loss: 0.6903 Iter 255, loss: 0.6678 Iter 256, loss: 0.6605 Iter 257, loss: 0.6516 Iter 258, loss: 0.6979 Iter 259, loss: 0.6454 Iter 260, loss: 0.6508 Iter 261, loss: 0.5896 Iter 262, loss: 0.6423 Iter 263, loss: 0.6989 Iter 264, loss: 0.6604 Iter 265, loss: 0.6183 Iter 266, loss: 0.6405 Iter 267, loss: 0.6294 Iter 268, loss: 0.6289 Iter 269, loss: 0.6212 Iter 270, loss: 0.6316 Iter 271, loss: 0.6050 Iter 272, loss: 0.6980 Iter 273, loss: 0.5966 Iter 274, loss: 0.6311 Iter 275, loss: 0.6385 Iter 276, loss: 0.6298 Iter 277, loss: 0.6493 Iter 278, loss: 0.6061 Iter 279, loss: 0.6007 Iter 280, loss: 0.6242 Iter 281, loss: 0.6410 Iter 282, loss: 0.6321 Iter 283, loss: 0.6623 Iter 284, loss: 0.6358 Iter 285, loss: 0.6360 Iter 286, loss: 0.6109 Iter 287, loss: 0.6384

Iter 288, loss: 0.5883 Iter 289, loss: 0.6315 Iter 290, loss: 0.6457 Iter 291, loss: 0.6526 Iter 292, loss: 0.6412 Iter 293, loss: 0.6680 Iter 294, loss: 0.6517 Iter 295, loss: 0.6186 Iter 296, loss: 0.6330 Iter 297, loss: 0.6469 Iter 298, loss: 0.6260 Iter 299, loss: 0.6048 Iter 300, loss: 0.6464 Iter 301, loss: 0.6536 Iter 302, loss: 0.6262 Iter 303, loss: 0.6517 Iter 304, loss: 0.7128 Iter 305, loss: 0.7030 Iter 306, loss: 0.6357 Iter 307, loss: 0.6451 Iter 308, loss: 0.6597 Iter 309, loss: 0.6387 Iter 310, loss: 0.6899 Iter 311, loss: 0.6719 Iter 312, loss: 0.6305 Iter 313, loss: 0.6554 Iter 314, loss: 0.6400 Iter 315, loss: 0.6329 Iter 316, loss: 0.6454 Iter 317, loss: 0.6125 Iter 318, loss: 0.6058 Iter 319, loss: 0.6636 Iter 320, loss: 0.6446 Iter 321, loss: 0.6516 Iter 322, loss: 0.6437 Iter 323, loss: 0.6604 Iter 324, loss: 0.6704 Iter 325, loss: 0.6281 Iter 326, loss: 0.6674 Iter 327, loss: 0.6026 Iter 328, loss: 0.6335 Iter 329, loss: 0.6347

Iter 330, loss: 0.6205 Iter 331, loss: 0.6626 Iter 332, loss: 0.6119 Iter 333, loss: 0.5783 Iter 334, loss: 0.6545 Iter 335, loss: 0.6175 Iter 336, loss: 0.6341 Iter 337, loss: 0.6580 Iter 338, loss: 0.6240 Iter 339, loss: 0.6710 Iter 340, loss: 0.6338 Iter 341, loss: 0.6168 Iter 342, loss: 0.6516 Iter 343, loss: 0.6571 Iter 344, loss: 0.6823 Iter 345, loss: 0.6177 Iter 346, loss: 0.6452 Iter 347, loss: 0.5935 Iter 348, loss: 0.6484 Iter 349, loss: 0.6488 Iter 350, loss: 0.6720 Iter 351, loss: 0.6367 Iter 352, loss: 0.6753 Iter 353, loss: 0.6008 Iter 354, loss: 0.6325 Iter 355, loss: 0.6557 Iter 356, loss: 0.6439 Iter 357, loss: 0.6733 Iter 358, loss: 0.6142 Iter 359, loss: 0.5907 Iter 360, loss: 0.6622 Iter 361, loss: 0.6562 Iter 362, loss: 0.6673 Iter 363, loss: 0.6186 Iter 364, loss: 0.6357 Iter 365, loss: 0.6232 Iter 366, loss: 0.6373 Iter 367, loss: 0.6309 Iter 368, loss: 0.6728 Iter 369, loss: 0.6801 Iter 370, loss: 0.6484

Iter 371, loss: 0.6354 Iter 372, loss: 0.6986 Iter 373, loss: 0.6493 Iter 374, loss: 0.6732 Iter 375, loss: 0.6207 Iter 376, loss: 0.6377 Iter 377, loss: 0.6827 Iter 378, loss: 0.6730 Iter 379, loss: 0.6371 Iter 380, loss: 0.6238 Iter 381, loss: 0.6284 Iter 382, loss: 0.6134 Iter 383, loss: 0.6458 Iter 384, loss: 0.6422 Iter 385, loss: 0.6133 Iter 386, loss: 0.6539 Iter 387, loss: 0.6348 Iter 388, loss: 0.6469 Iter 389, loss: 0.6385 Iter 390, loss: 0.6394 Iter 391, loss: 0.6686 Iter 392, loss: 0.6274 Iter 393, loss: 0.6292 Iter 394, loss: 0.6468 Iter 395, loss: 0.6288 Iter 396, loss: 0.6059 Iter 397, loss: 0.6721 Iter 398, loss: 0.6451 Iter 399, loss: 0.6520 Iter 400, loss: 0.6361 Iter 401, loss: 0.6849 Iter 402, loss: 0.6006 Iter 403, loss: 0.6501 Iter 404, loss: 0.6210 Iter 405, loss: 0.6418 Iter 406, loss: 0.6591 Iter 407, loss: 0.6562 Iter 408, loss: 0.6569 Iter 409, loss: 0.7001 Iter 410, loss: 0.6418 Iter 411, loss: 0.6460 Iter 412, loss: 0.6361

Iter 413, loss: 0.6627 Iter 414, loss: 0.6607 Iter 415, loss: 0.6543 Iter 416, loss: 0.6329 Iter 417, loss: 0.6272 Iter 418, loss: 0.6831 Iter 419, loss: 0.6587 Iter 420, loss: 0.6111 Iter 421, loss: 0.6316 Iter 422, loss: 0.5967 Iter 423, loss: 0.6394 Iter 424, loss: 0.6555 Iter 425, loss: 0.5965 Iter 426, loss: 0.6051 Iter 427, loss: 0.6989 Iter 428, loss: 0.6072 Iter 429, loss: 0.6405 Iter 430, loss: 0.6011 Iter 431, loss: 0.6317 Iter 432, loss: 0.6330 Iter 433, loss: 0.6688 Iter 434, loss: 0.6098 Iter 435, loss: 0.6869 Iter 436, loss: 0.6301 Iter 437, loss: 0.6423 Iter 438, loss: 0.5856 Iter 439, loss: 0.6614 Iter 440, loss: 0.6264 Iter 441, loss: 0.6459 Iter 442, loss: 0.6334 Iter 443, loss: 0.6682 Iter 444, loss: 0.6451 Iter 445, loss: 0.6336 Iter 446, loss: 0.6263 Iter 447, loss: 0.6782 Iter 448, loss: 0.6213 Iter 449, loss: 0.6114 Iter 450, loss: 0.6685 Iter 451, loss: 0.6227 Iter 452, loss: 0.6093 Iter 453, loss: 0.6869 Iter 454, loss: 0.5951

Iter 455, loss: 0.6397 Iter 456, loss: 0.6304 Iter 457, loss: 0.6266 Iter 458, loss: 0.5992 Iter 459, loss: 0.6282 Iter 460, loss: 0.6120 Iter 461, loss: 0.7017 Iter 462, loss: 0.5972 Iter 463, loss: 0.6447 Iter 464, loss: 0.6016 Iter 465, loss: 0.6475 Iter 466, loss: 0.6596 Iter 467, loss: 0.5907 Iter 468, loss: 0.6379 Iter 469, loss: 0.6863 Iter 470, loss: 0.6416 Iter 471, loss: 0.6296 Iter 472, loss: 0.6284 Iter 473, loss: 0.6412 Iter 474, loss: 0.6202 Iter 475, loss: 0.6682 Iter 476, loss: 0.6578 Iter 477, loss: 0.6336 Iter 478, loss: 0.6160 Iter 479, loss: 0.6263 Iter 480, loss: 0.6633 Iter 481, loss: 0.5978 Iter 482, loss: 0.6593 Iter 483, loss: 0.6035 Iter 484, loss: 0.6295 Iter 485, loss: 0.6378 Iter 486, loss: 0.6532 Iter 487, loss: 0.6849 Iter 488, loss: 0.6085 Iter 489, loss: 0.6268 Iter 490, loss: 0.5874 Iter 491, loss: 0.6172 Iter 492, loss: 0.6050 Iter 493, loss: 0.5907 Iter 494, loss: 0.6495 Iter 495, loss: 0.6440

Iter 496, loss: 0.6499 Iter 497, loss: 0.5843 Iter 498, loss: 0.6359 Iter 499, loss: 0.6503 Iter 500, loss: 0.6795 Iter 501, loss: 0.6226 Iter 502, loss: 0.6829 Iter 503, loss: 0.6492 Iter 504, loss: 0.6351 Iter 505, loss: 0.6418 Iter 506, loss: 0.6180 Iter 507, loss: 0.5617 Iter 508, loss: 0.6302 Iter 509, loss: 0.6408 Iter 510, loss: 0.6268 Iter 511, loss: 0.6852 Iter 512, loss: 0.6791 Iter 513, loss: 0.6861 Iter 514, loss: 0.6607 Iter 515, loss: 0.6385 Iter 516, loss: 0.6890 Iter 517, loss: 0.6292 Iter 518, loss: 0.5869 Iter 519, loss: 0.6299 Iter 520, loss: 0.6107 Iter 521, loss: 0.6594 Iter 522, loss: 0.6429 Iter 523, loss: 0.6706 Iter 524, loss: 0.6336 Iter 525, loss: 0.5998 Iter 526, loss: 0.6287 Iter 527, loss: 0.6561 Iter 528, loss: 0.6687 Iter 529, loss: 0.6021 Iter 530, loss: 0.5789 Iter 531, loss: 0.6445 Iter 532, loss: 0.6411 Iter 533, loss: 0.6666 Iter 534, loss: 0.6014 Iter 535, loss: 0.5940 Iter 536, loss: 0.6235 Iter 537, loss: 0.6249

Iter 538, loss: 0.6514 Iter 539, loss: 0.6297 Iter 540, loss: 0.5854 Iter 541, loss: 0.5765 Iter 542, loss: 0.5965 Iter 543, loss: 0.5760 Iter 544, loss: 0.6750 Iter 545, loss: 0.6423 Iter 546, loss: 0.6518 Iter 547, loss: 0.6292 Iter 548, loss: 0.6922 Iter 549, loss: 0.6803 Iter 550, loss: 0.6140 Iter 551, loss: 0.6387 Iter 552, loss: 0.6785 Iter 553, loss: 0.6517 Iter 554, loss: 0.5957 Iter 555, loss: 0.6492 Iter 556, loss: 0.6724 Iter 557, loss: 0.6550 Iter 558, loss: 0.6468 Iter 559, loss: 0.6720 Iter 560, loss: 0.6259 Iter 561, loss: 0.6222 Iter 562, loss: 0.6488 Iter 563, loss: 0.6254 Iter 564, loss: 0.6772 Iter 565, loss: 0.6016 Iter 566, loss: 0.6119 Iter 567, loss: 0.6481 Iter 568, loss: 0.6326 Iter 569, loss: 0.6431 Iter 570, loss: 0.6489 Iter 571, loss: 0.5973 Iter 572, loss: 0.6399 Iter 573, loss: 0.6323 Iter 574, loss: 0.6321 Iter 575, loss: 0.6579 Iter 576, loss: 0.6102 Iter 577, loss: 0.6306 Iter 578, loss: 0.7164 Iter 579, loss: 0.6533

Iter 580, loss: 0.6441 Iter 581, loss: 0.6315 Iter 582, loss: 0.5976 Iter 583, loss: 0.5975 Iter 584, loss: 0.6334 Iter 585, loss: 0.6143 Iter 586, loss: 0.6502 Iter 587, loss: 0.5678 Iter 588, loss: 0.6767 Iter 589, loss: 0.6186 Iter 590, loss: 0.5854 Iter 591, loss: 0.6506 Iter 592, loss: 0.6360 Iter 593, loss: 0.6311 Iter 594, loss: 0.6803 Iter 595, loss: 0.6325 Iter 596, loss: 0.6527 Iter 597, loss: 0.6180 Iter 598, loss: 0.6562 Iter 599, loss: 0.6343 Iter 600, loss: 0.6260 Iter 601, loss: 0.6493 Iter 602, loss: 0.6309 Iter 603, loss: 0.6267 Iter 604, loss: 0.6256 Iter 605, loss: 0.6000 Iter 606, loss: 0.5947 Iter 607, loss: 0.6528 Iter 608, loss: 0.6433 Iter 609, loss: 0.7294 Iter 610, loss: 0.6049 Iter 611, loss: 0.6538 Iter 612, loss: 0.6481 Iter 613, loss: 0.6186 Iter 614, loss: 0.6734 Iter 615, loss: 0.6214 Iter 616, loss: 0.6442 Iter 617, loss: 0.6221 Iter 618, loss: 0.6453 Iter 619, loss: 0.6195 Iter 620, loss: 0.6217 Iter 621, loss: 0.6099

Iter 622, loss: 0.6363 Iter 623, loss: 0.6630 Iter 624, loss: 0.6167 Iter 625, loss: 0.6159 Iter 626, loss: 0.6457 Iter 627, loss: 0.6719 Iter 628, loss: 0.6284 Iter 629, loss: 0.6118 Iter 630, loss: 0.6155 Iter 631, loss: 0.5841 Iter 632, loss: 0.6402 Iter 633, loss: 0.6494 Iter 634, loss: 0.6319 Iter 635, loss: 0.6424 Iter 636, loss: 0.6700 Iter 637, loss: 0.6106 Iter 638, loss: 0.6132 Iter 639, loss: 0.6696 Iter 640, loss: 0.6387 Iter 641, loss: 0.6416 Iter 642, loss: 0.5952 Iter 643, loss: 0.6510 Iter 644, loss: 0.6401 Iter 645, loss: 0.6389 Iter 646, loss: 0.6438 Iter 647, loss: 0.6546 Iter 648, loss: 0.6248 Iter 649, loss: 0.6340 Iter 650, loss: 0.6260 Iter 651, loss: 0.6554 Iter 652, loss: 0.6345 Iter 653, loss: 0.6484 Iter 654, loss: 0.6176 Iter 655, loss: 0.6612 Iter 656, loss: 0.6323 Iter 657, loss: 0.6525 Iter 658, loss: 0.6302 Iter 659, loss: 0.6232 Iter 660, loss: 0.6332 Iter 661, loss: 0.6174 Iter 662, loss: 0.6628 Iter 663, loss: 0.6449

Iter 664, loss: 0.6393 Iter 665, loss: 0.5962 Iter 666, loss: 0.6542 Iter 667, loss: 0.6780 Iter 668, loss: 0.6210 Iter 669, loss: 0.6231 Iter 670, loss: 0.6057 Iter 671, loss: 0.5786 Iter 672, loss: 0.6182 Iter 673, loss: 0.6135 Iter 674, loss: 0.6645 Iter 675, loss: 0.6328 Iter 676, loss: 0.6408 Iter 677, loss: 0.6738 Iter 678, loss: 0.6492 Iter 679, loss: 0.6069 Iter 680, loss: 0.6411 Iter 681, loss: 0.6301 Iter 682, loss: 0.6174 Iter 683, loss: 0.6754 Iter 684, loss: 0.6631 Iter 685, loss: 0.6585 Iter 686, loss: 0.6127 Iter 687, loss: 0.6552 Iter 688, loss: 0.6361 Iter 689, loss: 0.6092 Iter 690, loss: 0.7215 Iter 691, loss: 0.6317 Iter 692, loss: 0.6567 Iter 693, loss: 0.7043 Iter 694, loss: 0.6333 Iter 695, loss: 0.6458 Iter 696, loss: 0.6289 Iter 697, loss: 0.6223 Iter 698, loss: 0.6333 Iter 699, loss: 0.6801 Iter 700, loss: 0.6414 Iter 701, loss: 0.6521 Iter 702, loss: 0.6626 Iter 703, loss: 0.6568 Iter 704, loss: 0.6751 Iter 705, loss: 0.6393

Iter 706, loss: 0.6734 Iter 707, loss: 0.6280 Iter 708, loss: 0.6287 Iter 709, loss: 0.5924 Iter 710, loss: 0.7031 Iter 711, loss: 0.6595 Iter 712, loss: 0.6418 Iter 713, loss: 0.6438 Iter 714, loss: 0.6449 Iter 715, loss: 0.6695 Iter 716, loss: 0.6132 Iter 717, loss: 0.6555 Iter 718, loss: 0.6711 Iter 719, loss: 0.6172 Iter 720, loss: 0.6302 Iter 721, loss: 0.6745 Iter 722, loss: 0.5835 Iter 723, loss: 0.5990 Iter 724, loss: 0.6355 Iter 725, loss: 0.6690 Iter 726, loss: 0.6845 Iter 727, loss: 0.6075 Iter 728, loss: 0.6677 Iter 729, loss: 0.6203 Iter 730, loss: 0.6383 Iter 731, loss: 0.6165 Iter 732, loss: 0.6436 Iter 733, loss: 0.6182 Iter 734, loss: 0.6066 Iter 735, loss: 0.6362 Iter 736, loss: 0.6166 Iter 737, loss: 0.5968 Iter 738, loss: 0.6475 Iter 739, loss: 0.6316 Iter 740, loss: 0.6269 Iter 741, loss: 0.6033 Iter 742, loss: 0.6436 Iter 743, loss: 0.7328 Iter 744, loss: 0.6149 Iter 745, loss: 0.6143 Iter 746, loss: 0.6228

Iter 747, loss: 0.6191 Iter 748, loss: 0.6789 Iter 749, loss: 0.6863 Iter 750, loss: 0.6824 Iter 751, loss: 0.6198 Iter 752, loss: 0.6227 Iter 753, loss: 0.6413 Iter 754, loss: 0.6275 Iter 755, loss: 0.6438 Iter 756, loss: 0.6340 Iter 757, loss: 0.6436 Iter 758, loss: 0.6185 Iter 759, loss: 0.6364 Iter 760, loss: 0.6667 Iter 761, loss: 0.6141 Iter 762, loss: 0.6804 Iter 763, loss: 0.6260 Iter 764, loss: 0.6645 Iter 765, loss: 0.6379 Iter 766, loss: 0.6082 Iter 767, loss: 0.6107 Iter 768, loss: 0.6231 Iter 769, loss: 0.6691 Iter 770, loss: 0.6568 Iter 771, loss: 0.6007 Iter 772, loss: 0.6247 Iter 773, loss: 0.6344 Iter 774, loss: 0.6331 Iter 775, loss: 0.6157 Iter 776, loss: 0.5919 Iter 777, loss: 0.5900 Iter 778, loss: 0.5956 Iter 779, loss: 0.6643 Iter 780, loss: 0.6437 Iter 781, loss: 0.6706 Iter 782, loss: 0.6622 Iter 783, loss: 0.6773 Iter 784, loss: 0.6291 Iter 785, loss: 0.6186 Iter 786, loss: 0.6777 Iter 787, loss: 0.5930 Iter 788, loss: 0.5901

Iter 789, loss: 0.6416 Iter 790, loss: 0.6368 Iter 791, loss: 0.6355 Iter 792, loss: 0.6253 Iter 793, loss: 0.6518 Iter 794, loss: 0.6190 Iter 795, loss: 0.6561 Iter 796, loss: 0.6217 Iter 797, loss: 0.6354 Iter 798, loss: 0.6520 Iter 799, loss: 0.6289 Iter 800, loss: 0.6464 Iter 801, loss: 0.6380 Iter 802, loss: 0.6869 Iter 803, loss: 0.6476 Iter 804, loss: 0.6178 Iter 805, loss: 0.6486 Iter 806, loss: 0.6219 Iter 807, loss: 0.6305 Iter 808, loss: 0.5924 Iter 809, loss: 0.6301 Iter 810, loss: 0.6052 Iter 811, loss: 0.6332 Iter 812, loss: 0.6147 Iter 813, loss: 0.6428 Iter 814, loss: 0.6050 Iter 815, loss: 0.6335 Iter 816, loss: 0.6028 Iter 817, loss: 0.6485 Iter 818, loss: 0.6044 Iter 819, loss: 0.6127 Iter 820, loss: 0.6637 Iter 821, loss: 0.6260 Iter 822, loss: 0.6233 Iter 823, loss: 0.6102 Iter 824, loss: 0.6335 Iter 825, loss: 0.5937 Iter 826, loss: 0.6360 Iter 827, loss: 0.6532 Iter 828, loss: 0.6471 Iter 829, loss: 0.5997

Iter 830, loss: 0.6602 Iter 831, loss: 0.6127 Iter 832, loss: 0.6589 Iter 833, loss: 0.6051 Iter 834, loss: 0.6037 Iter 835, loss: 0.6537 Iter 836, loss: 0.5819 Iter 837, loss: 0.6114 Iter 838, loss: 0.6430 Iter 839, loss: 0.6647 Iter 840, loss: 0.6195 Iter 841, loss: 0.6024 Iter 842, loss: 0.6357 Iter 843, loss: 0.6010 Iter 844, loss: 0.6807 Iter 845, loss: 0.6495 Iter 846, loss: 0.6553 Iter 847, loss: 0.6602 Iter 848, loss: 0.6538 Iter 849, loss: 0.6167 Iter 850, loss: 0.6317 Iter 851, loss: 0.6572 Iter 852, loss: 0.5987 Iter 853, loss: 0.6414 Iter 854, loss: 0.6522 Iter 855, loss: 0.6159 Iter 856, loss: 0.6154 Iter 857, loss: 0.6379 Iter 858, loss: 0.6571 Iter 859, loss: 0.6000 Iter 860, loss: 0.6520 Iter 861, loss: 0.6613 Iter 862, loss: 0.6539 Iter 863, loss: 0.6945 Iter 864, loss: 0.5885 Iter 865, loss: 0.6163 Iter 866, loss: 0.6321 Iter 867, loss: 0.6125 Iter 868, loss: 0.6503 Iter 869, loss: 0.6360 Iter 870, loss: 0.6445 Iter 871, loss: 0.6455

Iter 872, loss: 0.6346 Iter 873, loss: 0.6358 Iter 874, loss: 0.5839 Iter 875, loss: 0.6343 Iter 876, loss: 0.6594 Iter 877, loss: 0.6074 Iter 878, loss: 0.5856 Iter 879, loss: 0.6038 Iter 880, loss: 0.6107 Iter 881, loss: 0.5340 Iter 882, loss: 0.6084 Iter 883, loss: 0.6483 Iter 884, loss: 0.6415 Iter 885, loss: 0.6433 Iter 886, loss: 0.6418 Iter 887, loss: 0.6443 Iter 888, loss: 0.6411 Iter 889, loss: 0.6087 Iter 890, loss: 0.6414 Iter 891, loss: 0.6525 Iter 892, loss: 0.6509 Iter 893, loss: 0.6596 Iter 894, loss: 0.5914 Iter 895, loss: 0.5847 Iter 896, loss: 0.6139 Iter 897, loss: 0.6372 Iter 898, loss: 0.6248 Iter 899, loss: 0.6387 Iter 900, loss: 0.6257 Iter 901, loss: 0.6804 Iter 902, loss: 0.5987 Iter 903, loss: 0.6378 Iter 904, loss: 0.6146 Iter 905, loss: 0.6165 Iter 906, loss: 0.6295 Iter 907, loss: 0.5887 Iter 908, loss: 0.6304 Iter 909, loss: 0.5952 Iter 910, loss: 0.6386 Iter 911, loss: 0.6444 Iter 912, loss: 0.6349 Iter 913, loss: 0.5484

Iter 914, loss: 0.6229 Iter 915, loss: 0.6242 Iter 916, loss: 0.6270 Iter 917, loss: 0.5573 Iter 918, loss: 0.6321 Iter 919, loss: 0.6204 Iter 920, loss: 0.6393 Iter 921, loss: 0.6373 Iter 922, loss: 0.6142 Iter 923, loss: 0.5980 Iter 924, loss: 0.6241 Iter 925, loss: 0.6348 Iter 926, loss: 0.6556 Iter 927, loss: 0.6227 Iter 928, loss: 0.6075 Iter 929, loss: 0.6140 Iter 930, loss: 0.5846 Iter 931, loss: 0.6030 Iter 932, loss: 0.6394 Iter 933, loss: 0.5968 Iter 934, loss: 0.6386 Iter 935, loss: 0.6764 Iter 936, loss: 0.6487 Iter 937, loss: 0.6480 Iter 938, loss: 0.6095 Iter 939, loss: 0.6237 Iter 940, loss: 0.6410 Iter 941, loss: 0.6492 Iter 942, loss: 0.5928 Iter 943, loss: 0.6444 Iter 944, loss: 0.6255 Iter 945, loss: 0.6491 Iter 946, loss: 0.6412 Iter 947, loss: 0.6873 Iter 948, loss: 0.6587 Iter 949, loss: 0.6234 Iter 950, loss: 0.6590 Iter 951, loss: 0.6609 Iter 952, loss: 0.6022 Iter 953, loss: 0.6106 Iter 954, loss: 0.6353 Iter 955, loss: 0.6031

Iter 956, loss: 0.6017 Iter 957, loss: 0.6575 Iter 958, loss: 0.6490 Iter 959, loss: 0.6145 Iter 960, loss: 0.6106 Iter 961, loss: 0.6549 Iter 962, loss: 0.6371 Iter 963, loss: 0.6325 Iter 964, loss: 0.6466 Iter 965, loss: 0.6618 Iter 966, loss: 0.6431 Iter 967, loss: 0.5900 Iter 968, loss: 0.6046 Iter 969, loss: 0.6043 Iter 970, loss: 0.5982 Iter 971, loss: 0.6117 Iter 972, loss: 0.5909 Iter 973, loss: 0.6246 Iter 974, loss: 0.6192 Iter 975, loss: 0.6153 Iter 976, loss: 0.6469 Iter 977, loss: 0.6368 Iter 978, loss: 0.6331 Iter 979, loss: 0.6258 Iter 980, loss: 0.6215 Iter 981, loss: 0.6068 Iter 982, loss: 0.6111 Iter 983, loss: 0.5840 Iter 984, loss: 0.6186 Iter 985, loss: 0.6721 Iter 986, loss: 0.6277 Iter 987, loss: 0.6179 Iter 988, loss: 0.6203 Iter 989, loss: 0.6687 Iter 990, loss: 0.6358 Iter 991, loss: 0.6355 Iter 992, loss: 0.6525 Iter 993, loss: 0.6059 Iter 994, loss: 0.5957 Iter 995, loss: 0.6409 Iter 996, loss: 0.6545 Iter 997, loss: 0.6463

Iter 998, loss: 0.6105 Iter 999, loss: 0.6648 Iter 1000, loss: 0.6594 Iter 1001, loss: 0.6531 Iter 1002, loss: 0.6240 Iter 1003, loss: 0.6282 Iter 1004, loss: 0.6646 Iter 1005, loss: 0.6426 Iter 1006, loss: 0.6167 Iter 1007, loss: 0.6066 Iter 1008, loss: 0.6228 Iter 1009, loss: 0.6660 Iter 1010, loss: 0.6641 Iter 1011, loss: 0.6407 Iter 1012, loss: 0.6568 Iter 1013, loss: 0.6481 Iter 1014, loss: 0.6686 Iter 1015, loss: 0.6109 Iter 1016, loss: 0.6337 Iter 1017, loss: 0.6491 Iter 1018, loss: 0.6098 Iter 1019, loss: 0.6174 Iter 1020, loss: 0.6070 Iter 1021, loss: 0.6189 Iter 1022, loss: 0.6010 Iter 1023, loss: 0.6697 Iter 1024, loss: 0.6474 Iter 1025, loss: 0.6181 Iter 1026, loss: 0.6301 Iter 1027, loss: 0.6347 Iter 1028, loss: 0.6206 Iter 1029, loss: 0.5878 Iter 1030, loss: 0.6265 Iter 1031, loss: 0.6113 Iter 1032, loss: 0.6508 Iter 1033, loss: 0.6058 Iter 1034, loss: 0.6533 Iter 1035, loss: 0.6426 Iter 1036, loss: 0.6056 Iter 1037, loss: 0.6319 Iter 1038, loss: 0.5863 Iter 1039, loss: 0.5929

Iter 1040, loss: 0.5761 Iter 1041, loss: 0.6235 Iter 1042, loss: 0.6428 Iter 1043, loss: 0.6484 Iter 1044, loss: 0.6351 Iter 1045, loss: 0.6369 Iter 1046, loss: 0.5990 Iter 1047, loss: 0.6388 Iter 1048, loss: 0.6182 Iter 1049, loss: 0.6129 Iter 1050, loss: 0.6116 Iter 1051, loss: 0.6167 Iter 1052, loss: 0.6657 Iter 1053, loss: 0.6802 Iter 1054, loss: 0.5778 Iter 1055, loss: 0.6609 Iter 1056, loss: 0.6212 Iter 1057, loss: 0.5584 Iter 1058, loss: 0.5995 Iter 1059, loss: 0.6393 Iter 1060, loss: 0.6349 Iter 1061, loss: 0.6045 Iter 1062, loss: 0.6401 Iter 1063, loss: 0.6116 Iter 1064, loss: 0.6042 Iter 1065, loss: 0.6580 Iter 1066, loss: 0.6648 Iter 1067, loss: 0.6661 Iter 1068, loss: 0.6153 Iter 1069, loss: 0.6540 Iter 1070, loss: 0.6123 Iter 1071, loss: 0.6289 Iter 1072, loss: 0.6559 Iter 1073, loss: 0.6511 Iter 1074, loss: 0.6131 Iter 1075, loss: 0.6552 Iter 1076, loss: 0.6427 Iter 1077, loss: 0.5852 Iter 1078, loss: 0.6368 Iter 1079, loss: 0.6189 Iter 1080, loss: 0.6387 Iter 1081, loss: 0.6190

Iter 1082, loss: 0.6170 Iter 1083, loss: 0.6136 Iter 1084, loss: 0.5934 Iter 1085, loss: 0.5760 Iter 1086, loss: 0.6147 Iter 1087, loss: 0.5961 Iter 1088, loss: 0.6704 Iter 1089, loss: 0.6304 Iter 1090, loss: 0.6314 Iter 1091, loss: 0.6105 Iter 1092, loss: 0.6218 Iter 1093, loss: 0.5762 Iter 1094, loss: 0.6301 Iter 1095, loss: 0.6062 Iter 1096, loss: 0.6291 Iter 1097, loss: 0.6950 Iter 1098, loss: 0.6086 Iter 1099, loss: 0.6356 Iter 1100, loss: 0.6105 Iter 1101, loss: 0.6073 Iter 1102, loss: 0.6609 Iter 1103, loss: 0.6404 Iter 1104, loss: 0.5798 Iter 1105, loss: 0.6191 Iter 1106, loss: 0.5913 Iter 1107, loss: 0.6292 Iter 1108, loss: 0.6566 Iter 1109, loss: 0.6019 Iter 1110, loss: 0.6767 Iter 1111, loss: 0.6426 Iter 1112, loss: 0.5827 Iter 1113, loss: 0.6246 Iter 1114, loss: 0.6462 Iter 1115, loss: 0.6601 Iter 1116, loss: 0.6469 Iter 1117, loss: 0.6118 Iter 1118, loss: 0.6261 Iter 1119, loss: 0.5948 Iter 1120, loss: 0.5852 Iter 1121, loss: 0.6136 Iter 1122, loss: 0.5925 Iter 1123, loss: 0.6602

Iter 1124, loss: 0.6368 Iter 1125, loss: 0.6072 Iter 1126, loss: 0.6220 Iter 1127, loss: 0.6183 Iter 1128, loss: 0.6157 Iter 1129, loss: 0.7042 Iter 1130, loss: 0.6657 Iter 1131, loss: 0.6116 Iter 1132, loss: 0.5925 Iter 1133, loss: 0.5816 Iter 1134, loss: 0.6394 Iter 1135, loss: 0.6284 Iter 1136, loss: 0.6633 Iter 1137, loss: 0.5855 Iter 1138, loss: 0.5914 Iter 1139, loss: 0.6631 Iter 1140, loss: 0.6038 Iter 1141, loss: 0.6425 Iter 1142, loss: 0.6355 Iter 1143, loss: 0.6163 Iter 1144, loss: 0.6140 Iter 1145, loss: 0.6031 Iter 1146, loss: 0.6263 Iter 1147, loss: 0.6373 Iter 1148, loss: 0.5903 Iter 1149, loss: 0.6220 Iter 1150, loss: 0.6052 Iter 1151, loss: 0.5694 Iter 1152, loss: 0.6338 Iter 1153, loss: 0.6282 Iter 1154, loss: 0.6325 Iter 1155, loss: 0.5798 Iter 1156, loss: 0.6829 Iter 1157, loss: 0.5720 Iter 1158, loss: 0.6819 Iter 1159, loss: 0.6434 Iter 1160, loss: 0.5844 Iter 1161, loss: 0.6482 Iter 1162, loss: 0.6385 Iter 1163, loss: 0.5961 Iter 1164, loss: 0.6448 Iter 1165, loss: 0.6122

Iter 1166, loss: 0.6162 Iter 1167, loss: 0.6522 Iter 1168, loss: 0.6356 Iter 1169, loss: 0.5546 Iter 1170, loss: 0.5956 Iter 1171, loss: 0.6469 Iter 1172, loss: 0.6084 Iter 1173, loss: 0.5734 Iter 1174, loss: 0.6177 Iter 1175, loss: 0.6173 Iter 1176, loss: 0.6601 Iter 1177, loss: 0.5884 Iter 1178, loss: 0.6259 Iter 1179, loss: 0.6871 Iter 1180, loss: 0.6584 Iter 1181, loss: 0.6571 Iter 1182, loss: 0.6486 Iter 1183, loss: 0.6110 Iter 1184, loss: 0.5969 Iter 1185, loss: 0.6182 Iter 1186, loss: 0.6166 Iter 1187, loss: 0.5978 Iter 1188, loss: 0.5774 Iter 1189, loss: 0.6775 Iter 1190, loss: 0.6231 Iter 1191, loss: 0.6620 Iter 1192, loss: 0.6373 Iter 1193, loss: 0.6343 Iter 1194, loss: 0.5702 Iter 1195, loss: 0.6609 Iter 1196, loss: 0.6159 Iter 1197, loss: 0.6060 Iter 1198, loss: 0.6566 Iter 1199, loss: 0.6545 Iter 1200, loss: 0.6190 Iter 1201, loss: 0.6179 Iter 1202, loss: 0.5992 Iter 1203, loss: 0.5899 Iter 1204, loss: 0.6125 Iter 1205, loss: 0.6350 Iter 1206, loss: 0.6343 Iter 1207, loss: 0.6472

Iter 1208, loss: 0.6432 Iter 1209, loss: 0.5749 Iter 1210, loss: 0.6009 Iter 1211, loss: 0.6229 Iter 1212, loss: 0.6038 Iter 1213, loss: 0.6352 Iter 1214, loss: 0.6175 Iter 1215, loss: 0.6200 Iter 1216, loss: 0.6093 Iter 1217, loss: 0.6285 Iter 1218, loss: 0.6373 Iter 1219, loss: 0.6181 Iter 1220, loss: 0.6590 Iter 1221, loss: 0.6567 Iter 1222, loss: 0.5980 Iter 1223, loss: 0.6550 Iter 1224, loss: 0.6423 Iter 1225, loss: 0.5942 Iter 1226, loss: 0.5860 Iter 1227, loss: 0.6800 Iter 1228, loss: 0.6305 Iter 1229, loss: 0.6831 Iter 1230, loss: 0.6208 Iter 1231, loss: 0.6782 Iter 1232, loss: 0.6073 Iter 1233, loss: 0.6057 Iter 1234, loss: 0.6392 Iter 1235, loss: 0.6798 Iter 1236, loss: 0.6210 Iter 1237, loss: 0.6412 Iter 1238, loss: 0.5665 Iter 1239, loss: 0.6124 Iter 1240, loss: 0.6229 Iter 1241, loss: 0.5790 Iter 1242, loss: 0.6089 Iter 1243, loss: 0.6306 Iter 1244, loss: 0.6257 Iter 1245, loss: 0.6520 Iter 1246, loss: 0.6449 Iter 1247, loss: 0.6180 Iter 1248, loss: 0.6349 Iter 1249, loss: 0.6469

Iter 1250, loss: 0.5746 Iter 1251, loss: 0.5967 Iter 1252, loss: 0.6358 Iter 1253, loss: 0.6259 Iter 1254, loss: 0.6022 Iter 1255, loss: 0.6271 Iter 1256, loss: 0.6642 Iter 1257, loss: 0.6194 Iter 1258, loss: 0.6417 Iter 1259, loss: 0.6509 Iter 1260, loss: 0.6338 Iter 1261, loss: 0.6065 Iter 1262, loss: 0.6418 Iter 1263, loss: 0.6497 Iter 1264, loss: 0.6280 Iter 1265, loss: 0.6185 Iter 1266, loss: 0.6226 Iter 1267, loss: 0.6307 Iter 1268, loss: 0.6350 Iter 1269, loss: 0.6296 Iter 1270, loss: 0.6453 Iter 1271, loss: 0.6242 Iter 1272, loss: 0.6158 Iter 1273, loss: 0.6264 Iter 1274, loss: 0.6025 Iter 1275, loss: 0.6077 Iter 1276, loss: 0.5965 Iter 1277, loss: 0.6433 Iter 1278, loss: 0.5971 Iter 1279, loss: 0.6365 Iter 1280, loss: 0.6201 Iter 1281, loss: 0.6391 Iter 1282, loss: 0.6164 Iter 1283, loss: 0.5982 Iter 1284, loss: 0.6138 Iter 1285, loss: 0.6288 Iter 1286, loss: 0.6259 Iter 1287, loss: 0.5904 Iter 1288, loss: 0.6567 Iter 1289, loss: 0.5878 Iter 1290, loss: 0.6220 Iter 1291, loss: 0.5534

Iter 1292, loss: 0.6345 Iter 1293, loss: 0.6282 Iter 1294, loss: 0.6096 Iter 1295, loss: 0.6167 Iter 1296, loss: 0.5972 Iter 1297, loss: 0.6351 Iter 1298, loss: 0.5652 Iter 1299, loss: 0.6150 Iter 1300, loss: 0.6462 Iter 1301, loss: 0.5944 Iter 1302, loss: 0.5797 Iter 1303, loss: 0.5786 Iter 1304, loss: 0.6078 Iter 1305, loss: 0.6590 Iter 1306, loss: 0.6152 Iter 1307, loss: 0.5944 Iter 1308, loss: 0.6405 Iter 1309, loss: 0.6257 Iter 1310, loss: 0.6165 Iter 1311, loss: 0.5947 Iter 1312, loss: 0.6423 Iter 1313, loss: 0.5932 Iter 1314, loss: 0.5781 Iter 1315, loss: 0.6821 Iter 1316, loss: 0.6199 Iter 1317, loss: 0.6833 Iter 1318, loss: 0.5879 Iter 1319, loss: 0.6567 Iter 1320, loss: 0.6328 Iter 1321, loss: 0.6146 Iter 1322, loss: 0.6172 Iter 1323, loss: 0.6097 Iter 1324, loss: 0.5949 Iter 1325, loss: 0.6031 Iter 1326, loss: 0.6001 Iter 1327, loss: 0.5825 Iter 1328, loss: 0.6013 Iter 1329, loss: 0.6321 Iter 1330, loss: 0.5958 Iter 1331, loss: 0.6529 Iter 1332, loss: 0.6540 Iter 1333, loss: 0.6208

Iter 1334, loss: 0.6174 Iter 1335, loss: 0.6045 Iter 1336, loss: 0.6408 Iter 1337, loss: 0.6428 Iter 1338, loss: 0.6487 Iter 1339, loss: 0.6272 Iter 1340, loss: 0.5770 Iter 1341, loss: 0.6202 Iter 1342, loss: 0.6465 Iter 1343, loss: 0.6398 Iter 1344, loss: 0.6542 Iter 1345, loss: 0.6147 Iter 1346, loss: 0.6482 Iter 1347, loss: 0.6167 Iter 1348, loss: 0.6119 Iter 1349, loss: 0.6536 Iter 1350, loss: 0.6205 Iter 1351, loss: 0.6146 Iter 1352, loss: 0.6486 Iter 1353, loss: 0.6352 Iter 1354, loss: 0.6324 Iter 1355, loss: 0.5768 Iter 1356, loss: 0.6362 Iter 1357, loss: 0.6358 Iter 1358, loss: 0.6302 Iter 1359, loss: 0.6029 Iter 1360, loss: 0.6303 Iter 1361, loss: 0.6703 Iter 1362, loss: 0.5805 Iter 1363, loss: 0.6425 Iter 1364, loss: 0.6406 Iter 1365, loss: 0.6271 Iter 1366, loss: 0.6552 Iter 1367, loss: 0.6646 Iter 1368, loss: 0.5322 Iter 1369, loss: 0.6175 Iter 1370, loss: 0.6145 Iter 1371, loss: 0.5861 Iter 1372, loss: 0.6356 Iter 1373, loss: 0.6003 Iter 1374, loss: 0.6059 Iter 1375, loss: 0.6575

Iter 1376, loss: 0.5884 Iter 1377, loss: 0.6225 Iter 1378, loss: 0.6168 Iter 1379, loss: 0.5918 Iter 1380, loss: 0.6230 Iter 1381, loss: 0.6413 Iter 1382, loss: 0.6020 Iter 1383, loss: 0.6203 Iter 1384, loss: 0.5898 Iter 1385, loss: 0.6056 Iter 1386, loss: 0.6427 Iter 1387, loss: 0.5923 Iter 1388, loss: 0.5708 Iter 1389, loss: 0.6062 Iter 1390, loss: 0.6279 Iter 1391, loss: 0.5733 Iter 1392, loss: 0.6177 Iter 1393, loss: 0.6059 Iter 1394, loss: 0.5471 Iter 1395, loss: 0.6339 Iter 1396, loss: 0.6083 Iter 1397, loss: 0.6290 Iter 1398, loss: 0.6352 Iter 1399, loss: 0.6391 Iter 1400, loss: 0.5987 Iter 1401, loss: 0.6608 Iter 1402, loss: 0.6316 Iter 1403, loss: 0.6180 Iter 1404, loss: 0.6404 Iter 1405, loss: 0.6523 Iter 1406, loss: 0.6328 Iter 1407, loss: 0.6134 Iter 1408, loss: 0.6235 Iter 1409, loss: 0.6735 Iter 1410, loss: 0.6013 Iter 1411, loss: 0.6142 Iter 1412, loss: 0.6494 Iter 1413, loss: 0.6588 Iter 1414, loss: 0.6303 Iter 1415, loss: 0.6289 Iter 1416, loss: 0.6704 Iter 1417, loss: 0.6388

Iter 1418, loss: 0.5650 Iter 1419, loss: 0.6333 Iter 1420, loss: 0.6175 Iter 1421, loss: 0.6227 Iter 1422, loss: 0.5674 Iter 1423, loss: 0.6071 Iter 1424, loss: 0.6226 Iter 1425, loss: 0.6415 Iter 1426, loss: 0.5812 Iter 1427, loss: 0.6471 Iter 1428, loss: 0.5815 Iter 1429, loss: 0.6304 Iter 1430, loss: 0.5982 Iter 1431, loss: 0.6420 Iter 1432, loss: 0.6777 Iter 1433, loss: 0.6141 Iter 1434, loss: 0.6390 Iter 1435, loss: 0.6486 Iter 1436, loss: 0.5931 Iter 1437, loss: 0.6222 Iter 1438, loss: 0.6066 Iter 1439, loss: 0.6023 Iter 1440, loss: 0.6444 Iter 1441, loss: 0.5952 Iter 1442, loss: 0.5990 Iter 1443, loss: 0.6652 Iter 1444, loss: 0.6071 Iter 1445, loss: 0.6330 Iter 1446, loss: 0.6080 Iter 1447, loss: 0.6453 Iter 1448, loss: 0.6100 Iter 1449, loss: 0.5968 Iter 1450, loss: 0.5988 Iter 1451, loss: 0.6227 Iter 1452, loss: 0.6184 Iter 1453, loss: 0.6597 Iter 1454, loss: 0.6119 Iter 1455, loss: 0.6215 Iter 1456, loss: 0.6185 Iter 1457, loss: 0.6570 Iter 1458, loss: 0.6292 Iter 1459, loss: 0.6474

Iter 1460, loss: 0.6272 Iter 1461, loss: 0.5986 Iter 1462, loss: 0.6422 Iter 1463, loss: 0.5864 Iter 1464, loss: 0.5738 Iter 1465, loss: 0.5929 Iter 1466, loss: 0.5982 Iter 1467, loss: 0.6345 Iter 1468, loss: 0.6362 Iter 1469, loss: 0.6230 Iter 1470, loss: 0.5977 Iter 1471, loss: 0.6235 Iter 1472, loss: 0.6355 Iter 1473, loss: 0.5862 Iter 1474, loss: 0.6438 Iter 1475, loss: 0.6778 Iter 1476, loss: 0.6397 Iter 1477, loss: 0.6010 Iter 1478, loss: 0.6548 Iter 1479, loss: 0.6449 Iter 1480, loss: 0.6070 Iter 1481, loss: 0.6302 Iter 1482, loss: 0.6381 Iter 1483, loss: 0.6019 Iter 1484, loss: 0.6058 Iter 1485, loss: 0.6164 Iter 1486, loss: 0.5800 Iter 1487, loss: 0.6081 Iter 1488, loss: 0.6485 Iter 1489, loss: 0.5989 Iter 1490, loss: 0.6329 Iter 1491, loss: 0.6145 Iter 1492, loss: 0.5977 Iter 1493, loss: 0.6176 Iter 1494, loss: 0.5844 Iter 1495, loss: 0.6172 Iter 1496, loss: 0.6135 Iter 1497, loss: 0.5945 Iter 1498, loss: 0.6370 Iter 1499, loss: 0.6404 Iter 1500, loss: 0.6010 Iter 1501, loss: 0.6356

Iter 1502, loss: 0.5837 Iter 1503, loss: 0.5990 Iter 1504, loss: 0.6114 Iter 1505, loss: 0.6354 Iter 1506, loss: 0.6711 Iter 1507, loss: 0.6432 Iter 1508, loss: 0.6215 Iter 1509, loss: 0.6204 Iter 1510, loss: 0.6591 Iter 1511, loss: 0.6163 Iter 1512, loss: 0.6106 Iter 1513, loss: 0.6519 Iter 1514, loss: 0.6540 Iter 1515, loss: 0.6611 Iter 1516, loss: 0.6198 Iter 1517, loss: 0.6001 Iter 1518, loss: 0.6523 Iter 1519, loss: 0.6643 Iter 1520, loss: 0.5978 Iter 1521, loss: 0.6639 Iter 1522, loss: 0.6467 Iter 1523, loss: 0.6244 Iter 1524, loss: 0.5908 Iter 1525, loss: 0.6313 Iter 1526, loss: 0.5848 Iter 1527, loss: 0.6710 Iter 1528, loss: 0.6195 Iter 1529, loss: 0.6460 Iter 1530, loss: 0.6085 Iter 1531, loss: 0.6430 Iter 1532, loss: 0.6152 Iter 1533, loss: 0.6275 Iter 1534, loss: 0.6405 Iter 1535, loss: 0.5903 Iter 1536, loss: 0.6336 Iter 1537, loss: 0.5834 Iter 1538, loss: 0.6052 Iter 1539, loss: 0.6402 Iter 1540, loss: 0.6189 Iter 1541, loss: 0.6559 Iter 1542, loss: 0.5983 Iter 1543, loss: 0.6193

Iter 1544, loss: 0.6342 Iter 1545, loss: 0.6470 Iter 1546, loss: 0.6467 Iter 1547, loss: 0.6194 Iter 1548, loss: 0.6051 Iter 1549, loss: 0.6140 Iter 1550, loss: 0.6058 Iter 1551, loss: 0.5915 Iter 1552, loss: 0.6587 Iter 1553, loss: 0.5882 Iter 1554, loss: 0.6276 Iter 1555, loss: 0.6308 Iter 1556, loss: 0.6518 Iter 1557, loss: 0.5790 Iter 1558, loss: 0.6497 Iter 1559, loss: 0.6385 Iter 1560, loss: 0.6210 Iter 1561, loss: 0.6163 Iter 1562, loss: 0.6152 Iter 1563, loss: 0.6543 Iter 1564, loss: 0.5779 Iter 1565, loss: 0.5967 Iter 1566, loss: 0.6194 Iter 1567, loss: 0.5810 Iter 1568, loss: 0.6178 Iter 1569, loss: 0.5993 Iter 1570, loss: 0.6569 Iter 1571, loss: 0.6693 Iter 1572, loss: 0.6511 Iter 1573, loss: 0.5856 Iter 1574, loss: 0.6310 Iter 1575, loss: 0.6073 Iter 1576, loss: 0.6444 Iter 1577, loss: 0.6362 Iter 1578, loss: 0.6378 Iter 1579, loss: 0.6252 Iter 1580, loss: 0.6083 Iter 1581, loss: 0.5904 Iter 1582, loss: 0.6441 Iter 1583, loss: 0.6114 Iter 1584, loss: 0.5872 Iter 1585, loss: 0.5814

Iter 1586, loss: 0.6227 Iter 1587, loss: 0.6448 Iter 1588, loss: 0.6183 Iter 1589, loss: 0.6238 Iter 1590, loss: 0.5990 Iter 1591, loss: 0.5812 Iter 1592, loss: 0.5868 Iter 1593, loss: 0.5704 Iter 1594, loss: 0.5941 Iter 1595, loss: 0.5903 Iter 1596, loss: 0.6146 Iter 1597, loss: 0.6157 Iter 1598, loss: 0.6077 Iter 1599, loss: 0.6664 Iter 1600, loss: 0.6366 Iter 1601, loss: 0.5951 Iter 1602, loss: 0.5894 Iter 1603, loss: 0.6864 Iter 1604, loss: 0.6345 Iter 1605, loss: 0.6440 Iter 1606, loss: 0.6217 Iter 1607, loss: 0.6291 Iter 1608, loss: 0.5988 Iter 1609, loss: 0.6192 Iter 1610, loss: 0.6230 Iter 1611, loss: 0.5894 Iter 1612, loss: 0.6743 Iter 1613, loss: 0.6003 Iter 1614, loss: 0.5914 Iter 1615, loss: 0.6402 Iter 1616, loss: 0.6208 Iter 1617, loss: 0.6057 Iter 1618, loss: 0.6041 Iter 1619, loss: 0.6723 Iter 1620, loss: 0.6068 Iter 1621, loss: 0.5903 Iter 1622, loss: 0.5526 Iter 1623, loss: 0.6092 Iter 1624, loss: 0.6359 Iter 1625, loss: 0.6285 Iter 1626, loss: 0.6323 Iter 1627, loss: 0.6746

Iter 1628, loss: 0.6281 Iter 1629, loss: 0.5940 Iter 1630, loss: 0.5898 Iter 1631, loss: 0.6255 Iter 1632, loss: 0.6358 Iter 1633, loss: 0.5832 Iter 1634, loss: 0.6003 Iter 1635, loss: 0.6290 Iter 1636, loss: 0.6492 Iter 1637, loss: 0.6107 Iter 1638, loss: 0.5687 Iter 1639, loss: 0.5936 Iter 1640, loss: 0.5826 Iter 1641, loss: 0.6253 Iter 1642, loss: 0.6149 Iter 1643, loss: 0.6894 Iter 1644, loss: 0.5733 Iter 1645, loss: 0.5830 Iter 1646, loss: 0.6199 Iter 1647, loss: 0.5724 Iter 1648, loss: 0.5760 Iter 1649, loss: 0.6274 Iter 1650, loss: 0.6445 Iter 1651, loss: 0.6495 Iter 1652, loss: 0.5936 Iter 1653, loss: 0.6200 Iter 1654, loss: 0.5726 Iter 1655, loss: 0.5609 Iter 1656, loss: 0.6298 Iter 1657, loss: 0.5841 Iter 1658, loss: 0.6086 Iter 1659, loss: 0.5925 Iter 1660, loss: 0.6490 Iter 1661, loss: 0.6460 Iter 1662, loss: 0.6135 Iter 1663, loss: 0.6200 Iter 1664, loss: 0.6491 Iter 1665, loss: 0.5763 Iter 1666, loss: 0.6394 Iter 1667, loss: 0.6554 Iter 1668, loss: 0.6253 Iter 1669, loss: 0.5896

Iter 1670, loss: 0.6433 Iter 1671, loss: 0.6006 Iter 1672, loss: 0.6023 Iter 1673, loss: 0.6531 Iter 1674, loss: 0.6188 Iter 1675, loss: 0.6555 Iter 1676, loss: 0.6197 Iter 1677, loss: 0.6462 Iter 1678, loss: 0.5901 Iter 1679, loss: 0.6089 Iter 1680, loss: 0.6027 Iter 1681, loss: 0.6089 Iter 1682, loss: 0.6144 Iter 1683, loss: 0.6336 Iter 1684, loss: 0.5728 Iter 1685, loss: 0.6099 Iter 1686, loss: 0.6355 Iter 1687, loss: 0.6384 Iter 1688, loss: 0.6227 Iter 1689, loss: 0.6358 Iter 1690, loss: 0.6118 Iter 1691, loss: 0.5879 Iter 1692, loss: 0.6307 Iter 1693, loss: 0.6268 Iter 1694, loss: 0.5851 Iter 1695, loss: 0.6215 Iter 1696, loss: 0.6707 Iter 1697, loss: 0.6519 Iter 1698, loss: 0.6015 Iter 1699, loss: 0.6206 Iter 1700, loss: 0.6602 Iter 1701, loss: 0.6147 Iter 1702, loss: 0.6075 Iter 1703, loss: 0.6142 Iter 1704, loss: 0.6364 Iter 1705, loss: 0.6067 Iter 1706, loss: 0.6158 Iter 1707, loss: 0.6670 Iter 1708, loss: 0.5991 Iter 1709, loss: 0.6199 Iter 1710, loss: 0.5897 Iter 1711, loss: 0.6009

Iter 1712, loss: 0.6122 Iter 1713, loss: 0.5706 Iter 1714, loss: 0.6257 Iter 1715, loss: 0.6604 Iter 1716, loss: 0.6018 Iter 1717, loss: 0.6845 Iter 1718, loss: 0.6554 Iter 1719, loss: 0.6144 Iter 1720, loss: 0.6106 Iter 1721, loss: 0.6087 Iter 1722, loss: 0.6353 Iter 1723, loss: 0.5936 Iter 1724, loss: 0.6629 Iter 1725, loss: 0.6352 Iter 1726, loss: 0.5802 Iter 1727, loss: 0.6010 Iter 1728, loss: 0.5967 Iter 1729, loss: 0.6005 Iter 1730, loss: 0.6150 Iter 1731, loss: 0.6036 Iter 1732, loss: 0.6221 Iter 1733, loss: 0.5981 Iter 1734, loss: 0.6039 Iter 1735, loss: 0.6508 Iter 1736, loss: 0.6055 Iter 1737, loss: 0.6229 Iter 1738, loss: 0.6742 Iter 1739, loss: 0.6114 Iter 1740, loss: 0.6137 Iter 1741, loss: 0.5775 Iter 1742, loss: 0.6097 Iter 1743, loss: 0.6016 Iter 1744, loss: 0.5945 Iter 1745, loss: 0.6151 Iter 1746, loss: 0.6456 Iter 1747, loss: 0.5869 Iter 1748, loss: 0.6402 Iter 1749, loss: 0.6152 Iter 1750, loss: 0.6050 Iter 1751, loss: 0.5746 Iter 1752, loss: 0.6335 Iter 1753, loss: 0.6292

Iter 1754, loss: 0.5788 Iter 1755, loss: 0.6492 Iter 1756, loss: 0.5995 Iter 1757, loss: 0.6211 Iter 1758, loss: 0.6209 Iter 1759, loss: 0.5825 Iter 1760, loss: 0.6241 Iter 1761, loss: 0.5816 Iter 1762, loss: 0.6126 Iter 1763, loss: 0.5987 Iter 1764, loss: 0.6118 Iter 1765, loss: 0.6072 Iter 1766, loss: 0.5973 Iter 1767, loss: 0.6113 Iter 1768, loss: 0.6204 Iter 1769, loss: 0.6402 Iter 1770, loss: 0.5908 Iter 1771, loss: 0.6172 Iter 1772, loss: 0.6399 Iter 1773, loss: 0.6546 Iter 1774, loss: 0.6316 Iter 1775, loss: 0.6448 Iter 1776, loss: 0.6059 Iter 1777, loss: 0.6215 Iter 1778, loss: 0.6226 Iter 1779, loss: 0.5911 Iter 1780, loss: 0.6315 Iter 1781, loss: 0.5952 Iter 1782, loss: 0.6341 Iter 1783, loss: 0.5884 Iter 1784, loss: 0.6488 Iter 1785, loss: 0.6172 Iter 1786, loss: 0.6111 Iter 1787, loss: 0.6269 Iter 1788, loss: 0.6350 Iter 1789, loss: 0.6298 Iter 1790, loss: 0.6019 Iter 1791, loss: 0.5844 Iter 1792, loss: 0.6335 Iter 1793, loss: 0.6317 Iter 1794, loss: 0.5810 Iter 1795, loss: 0.6172

Iter 1796, loss: 0.6060 Iter 1797, loss: 0.6170 Iter 1798, loss: 0.6272 Iter 1799, loss: 0.6035 Iter 1800, loss: 0.6440 Iter 1801, loss: 0.6309 Iter 1802, loss: 0.6276 Iter 1803, loss: 0.6117 Iter 1804, loss: 0.6617 Iter 1805, loss: 0.6457 Iter 1806, loss: 0.6028 Iter 1807, loss: 0.6350 Iter 1808, loss: 0.5823 Iter 1809, loss: 0.6094 Iter 1810, loss: 0.6403 Iter 1811, loss: 0.6105 Iter 1812, loss: 0.6426 Iter 1813, loss: 0.5935 Iter 1814, loss: 0.6543 Iter 1815, loss: 0.5884 Iter 1816, loss: 0.6266 Iter 1817, loss: 0.6642 Iter 1818, loss: 0.5980 Iter 1819, loss: 0.6548 Iter 1820, loss: 0.6023 Iter 1821, loss: 0.6524 Iter 1822, loss: 0.6359 Iter 1823, loss: 0.6403 Iter 1824, loss: 0.6050 Iter 1825, loss: 0.6447 Iter 1826, loss: 0.6253 Iter 1827, loss: 0.5791 Iter 1828, loss: 0.5840 Iter 1829, loss: 0.6460 Iter 1830, loss: 0.6602 Iter 1831, loss: 0.5968 Iter 1832, loss: 0.6621 Iter 1833, loss: 0.5918 Iter 1834, loss: 0.6203 Iter 1835, loss: 0.6040 Iter 1836, loss: 0.6137 Iter 1837, loss: 0.6023

Iter 1838, loss: 0.5608 Iter 1839, loss: 0.5936 Iter 1840, loss: 0.6371 Iter 1841, loss: 0.5927 Iter 1842, loss: 0.6364 Iter 1843, loss: 0.5845 Iter 1844, loss: 0.6231 Iter 1845, loss: 0.5715 Iter 1846, loss: 0.5532 Iter 1847, loss: 0.6238 Iter 1848, loss: 0.6437 Iter 1849, loss: 0.5918 Iter 1850, loss: 0.5964 Iter 1851, loss: 0.6214 Iter 1852, loss: 0.5701 Iter 1853, loss: 0.6101 Iter 1854, loss: 0.6172 Iter 1855, loss: 0.6194 Iter 1856, loss: 0.6311 Iter 1857, loss: 0.6170 Iter 1858, loss: 0.6558 Iter 1859, loss: 0.6420

Iter 1860, loss: 0.5966 Iter 1861, loss: 0.5983 Iter 1862, loss: 0.5953 Iter 1863, loss: 0.6107 Iter 1864, loss: 0.6305 Iter 1865, loss: 0.6191 Iter 1866, loss: 0.6463 Iter 1867, loss: 0.5830 Iter 1868, loss: 0.5910 Iter 1869, loss: 0.6004 Iter 1870, loss: 0.6629 Iter 1871, loss: 0.6122 Iter 1872, loss: 0.5979 Iter 1873, loss: 0.6150 Iter 1874, loss: 0.6063 Iter 1875, loss: 0.6582 Iter 1876, loss: 0.5699 Iter 1877, loss: 0.6005 Iter 1878, loss: 0.6053 Iter 1879, loss: 0.6362 Iter 1880, loss: 0.6541 Iter 1881, loss: 0.5998 Iter 1882, loss: 0.6486 Iter 1883, loss: 0.6142 Iter 1884, loss: 0.6151 Iter 1885, loss: 0.6444 Iter 1886, loss: 0.5966 Iter 1887, loss: 0.6117 Iter 1888, loss: 0.5839 Iter 1889, loss: 0.6214 Iter 1890, loss: 0.6293 Iter 1891, loss: 0.6380 Iter 1892, loss: 0.6008 Iter 1893, loss: 0.5883 Iter 1894, loss: 0.5979 Iter 1895, loss: 0.6393 Iter 1896, loss: 0.6028 Iter 1897, loss: 0.6415 Iter 1898, loss: 0.6494 Iter 1899, loss: 0.6292 Iter 1900, loss: 0.6405

Iter 1901, loss: 0.6168 Iter 1902, loss: 0.5817 Iter 1903, loss: 0.6297 Iter 1904, loss: 0.6004 Iter 1905, loss: 0.5418 Iter 1906, loss: 0.6197 Iter 1907, loss: 0.6139 Iter 1908, loss: 0.5864 Iter 1909, loss: 0.5660 Iter 1910, loss: 0.6024 Iter 1911, loss: 0.6841 Iter 1912, loss: 0.5951 Iter 1913, loss: 0.6452 Iter 1914, loss: 0.5519 Iter 1915, loss: 0.5910 Iter 1916, loss: 0.5879 Iter 1917, loss: 0.5994 Iter 1918, loss: 0.5967 Iter 1919, loss: 0.6075 Iter 1920, loss: 0.6417 Iter 1921, loss: 0.5905 Iter 1922, loss: 0.6474 Iter 1923, loss: 0.6063 Iter 1924, loss: 0.5857 Iter 1925, loss: 0.6401 Iter 1926, loss: 0.6008 Iter 1927, loss: 0.6201 Iter 1928, loss: 0.6270 Iter 1929, loss: 0.6053 Iter 1930, loss: 0.6180 Iter 1931, loss: 0.6465 Iter 1932, loss: 0.6545 Iter 1933, loss: 0.6034 Iter 1934, loss: 0.6287 Iter 1935, loss: 0.6025 Iter 1936, loss: 0.6563 Iter 1937, loss: 0.5784 Iter 1938, loss: 0.5939 Iter 1939, loss: 0.6027 Iter 1940, loss: 0.6246 Iter 1941, loss: 0.6332

Iter 1942, loss: 0.6002 Iter 1943, loss: 0.6250 Iter 1944, loss: 0.6493 Iter 1945, loss: 0.5891 Iter 1946, loss: 0.5962 Iter 1947, loss: 0.5874 Iter 1948, loss: 0.6417 Iter 1949, loss: 0.6121 Iter 1950, loss: 0.6379 Iter 1951, loss: 0.6206 Iter 1952, loss: 0.5951 Iter 1953, loss: 0.5795 Iter 1954, loss: 0.6255 Iter 1955, loss: 0.6283 Iter 1956, loss: 0.6575 Iter 1957, loss: 0.6228 Iter 1958, loss: 0.6212 Iter 1959, loss: 0.6030 Iter 1960, loss: 0.6355 Iter 1961, loss: 0.6187 Iter 1962, loss: 0.6195 Iter 1963, loss: 0.6246 Iter 1964, loss: 0.6545 Iter 1965, loss: 0.6354 Iter 1966, loss: 0.6173 Iter 1967, loss: 0.6367 Iter 1968, loss: 0.6013 Iter 1969, loss: 0.5942 Iter 1970, loss: 0.6243 Iter 1971, loss: 0.5793 Iter 1972, loss: 0.5929 Iter 1973, loss: 0.5609 Iter 1974, loss: 0.6002 Iter 1975, loss: 0.5768 Iter 1976, loss: 0.5766 Iter 1977, loss: 0.6854 Iter 1978, loss: 0.5891 Iter 1979, loss: 0.6103 Iter 1980, loss: 0.6200 Iter 1981, loss: 0.5794 Iter 1982, loss: 0.6049

Iter 1983, loss: 0.6104 Iter 1984, loss: 0.6110 Iter 1985, loss: 0.6359 Iter 1986, loss: 0.6220 Iter 1987, loss: 0.6378 Iter 1988, loss: 0.6275 Iter 1989, loss: 0.6374 Iter 1990, loss: 0.5775 Iter 1991, loss: 0.6432 Iter 1992, loss: 0.5813 Iter 1993, loss: 0.5932 Iter 1994, loss: 0.6257 Iter 1995, loss: 0.6762 Iter 1996, loss: 0.6046 Iter 1997, loss: 0.6315 Iter 1998, loss: 0.6129 Iter 1999, loss: 0.5949 Iter 2000, loss: 0.5952 Iter 2001, loss: 0.6544 Iter 2002, loss: 0.6708 Iter 2003, loss: 0.6263 Iter 2004, loss: 0.6157 Iter 2005, loss: 0.5832 Iter 2006, loss: 0.5829 Iter 2007, loss: 0.6281 Iter 2008, loss: 0.6326 Iter 2009, loss: 0.6101 Iter 2010, loss: 0.5914 Iter 2011, loss: 0.6692 Iter 2012, loss: 0.6114 Iter 2013, loss: 0.6144 Iter 2014, loss: 0.6358 Iter 2015, loss: 0.6273 Iter 2016, loss: 0.6053 Iter 2017, loss: 0.6406 Iter 2018, loss: 0.6241 Iter 2019, loss: 0.6155 Iter 2020, loss: 0.6254 Iter 2021, loss: 0.6175 Iter 2022, loss: 0.5797 Iter 2023, loss: 0.6224

Iter 2024, loss: 0.6155 Iter 2025, loss: 0.5904 Iter 2026, loss: 0.6037 Iter 2027, loss: 0.5994 Iter 2028, loss: 0.6155 Iter 2029, loss: 0.5976 Iter 2030, loss: 0.6020 Iter 2031, loss: 0.6207 Iter 2032, loss: 0.5843 Iter 2033, loss: 0.5924 Iter 2034, loss: 0.6351 Iter 2035, loss: 0.6060 Iter 2036, loss: 0.5958 Iter 2037, loss: 0.5780 Iter 2038, loss: 0.5826 Iter 2039, loss: 0.6072 Iter 2040, loss: 0.5837 Iter 2041, loss: 0.6344 Iter 2042, loss: 0.5902 Iter 2043, loss: 0.6514 Iter 2044, loss: 0.6170 Iter 2045, loss: 0.5869 Iter 2046, loss: 0.5979 Iter 2047, loss: 0.6078 Iter 2048, loss: 0.5985 Iter 2049, loss: 0.6108 Iter 2050, loss: 0.6371 Iter 2051, loss: 0.6284 Iter 2052, loss: 0.6535 Iter 2053, loss: 0.6461 Iter 2054, loss: 0.6260 Iter 2055, loss: 0.6196 Iter 2056, loss: 0.6100 Iter 2057, loss: 0.5906 Iter 2058, loss: 0.6190 Iter 2059, loss: 0.6180 Iter 2060, loss: 0.6176 Iter 2061, loss: 0.6099 Iter 2062, loss: 0.6016 Iter 2063, loss: 0.5901 Iter 2064, loss: 0.6306 Iter 2065, loss: 0.5687

Iter 2066, loss: 0.6315 Iter 2067, loss: 0.6247 Iter 2068, loss: 0.5867 Iter 2069, loss: 0.6133 Iter 2070, loss: 0.6231 Iter 2071, loss: 0.6277 Iter 2072, loss: 0.6211 Iter 2073, loss: 0.6428 Iter 2074, loss: 0.5879 Iter 2075, loss: 0.5704 Iter 2076, loss: 0.5798 Iter 2077, loss: 0.5975 Iter 2078, loss: 0.6328 Iter 2079, loss: 0.6169 Iter 2080, loss: 0.6383 Iter 2081, loss: 0.6177 Iter 2082, loss: 0.5690 Iter 2083, loss: 0.5562 Iter 2084, loss: 0.5783 Iter 2085, loss: 0.6382 Iter 2086, loss: 0.6121 Iter 2087, loss: 0.6183 Iter 2088, loss: 0.5687 Iter 2089, loss: 0.6095 Iter 2090, loss: 0.5560 Iter 2091, loss: 0.6407 Iter 2092, loss: 0.6476 Iter 2093, loss: 0.5921 Iter 2094, loss: 0.5988 Iter 2095, loss: 0.6349 Iter 2096, loss: 0.5873 Iter 2097, loss: 0.6349 Iter 2098, loss: 0.6252 Iter 2099, loss: 0.6135 Iter 2100, loss: 0.6688 Iter 2101, loss: 0.5716 Iter 2102, loss: 0.5741 Iter 2103, loss: 0.5988 Iter 2104, loss: 0.5936 Iter 2105, loss: 0.6609 Iter 2106, loss: 0.5780

Iter 2107, loss: 0.6076 Iter 2108, loss: 0.6149 Iter 2109, loss: 0.5880 Iter 2110, loss: 0.6205 Iter 2111, loss: 0.6034 Iter 2112, loss: 0.6092 Iter 2113, loss: 0.5951 Iter 2114, loss: 0.6347 Iter 2115, loss: 0.6020 Iter 2116, loss: 0.6103 Iter 2117, loss: 0.6284 Iter 2118, loss: 0.6252 Iter 2119, loss: 0.6452 Iter 2120, loss: 0.6290 Iter 2121, loss: 0.6366 Iter 2122, loss: 0.5850 Iter 2123, loss: 0.6173 Iter 2124, loss: 0.5797 Iter 2125, loss: 0.5903 Iter 2126, loss: 0.5860 Iter 2127, loss: 0.5788 Iter 2128, loss: 0.6072 Iter 2129, loss: 0.5828 Iter 2130, loss: 0.6200 Iter 2131, loss: 0.6029 Iter 2132, loss: 0.5970 Iter 2133, loss: 0.5892 Iter 2134, loss: 0.6659 Iter 2135, loss: 0.6511 Iter 2136, loss: 0.6266 Iter 2137, loss: 0.6427 Iter 2138, loss: 0.5894 Iter 2139, loss: 0.6165 Iter 2140, loss: 0.6666 Iter 2141, loss: 0.6290 Iter 2142, loss: 0.6515 Iter 2143, loss: 0.6131 Iter 2144, loss: 0.6039 Iter 2145, loss: 0.6043 Iter 2146, loss: 0.6129 Iter 2147, loss: 0.6050 Iter 2148, loss: 0.6130

Iter 2149, loss: 0.6339 Iter 2150, loss: 0.6557 Iter 2151, loss: 0.6149 Iter 2152, loss: 0.5976 Iter 2153, loss: 0.5859 Iter 2154, loss: 0.5853 Iter 2155, loss: 0.5842 Iter 2156, loss: 0.6039 Iter 2157, loss: 0.6259 Iter 2158, loss: 0.6501 Iter 2159, loss: 0.6713 Iter 2160, loss: 0.6497 Iter 2161, loss: 0.5834 Iter 2162, loss: 0.6164 Iter 2163, loss: 0.6467 Iter 2164, loss: 0.6500 Iter 2165, loss: 0.5981 Iter 2166, loss: 0.6064 Iter 2167, loss: 0.6336 Iter 2168, loss: 0.6229 Iter 2169, loss: 0.5768 Iter 2170, loss: 0.5507 Iter 2171, loss: 0.5842 Iter 2172, loss: 0.6269 Iter 2173, loss: 0.6231 Iter 2174, loss: 0.6553 Iter 2175, loss: 0.5880 Iter 2176, loss: 0.6190 Iter 2177, loss: 0.5640 Iter 2178, loss: 0.6377 Iter 2179, loss: 0.6147 Iter 2180, loss: 0.5995 Iter 2181, loss: 0.6175 Iter 2182, loss: 0.5653 Iter 2183, loss: 0.5851 Iter 2184, loss: 0.6444 Iter 2185, loss: 0.5735 Iter 2186, loss: 0.6227 Iter 2187, loss: 0.6372 Iter 2188, loss: 0.6123 Iter 2189, loss: 0.6408

Iter 2190, loss: 0.6197 Iter 2191, loss: 0.6388 Iter 2192, loss: 0.5800 Iter 2193, loss: 0.6186 Iter 2194, loss: 0.6235 Iter 2195, loss: 0.6244 Iter 2196, loss: 0.6214 Iter 2197, loss: 0.5852 Iter 2198, loss: 0.6053 Iter 2199, loss: 0.5812 Iter 2200, loss: 0.6336 Iter 2201, loss: 0.6096 Iter 2202, loss: 0.6190 Iter 2203, loss: 0.6271 Iter 2204, loss: 0.6234 Iter 2205, loss: 0.6216 Iter 2206, loss: 0.5604 Iter 2207, loss: 0.6642 Iter 2208, loss: 0.6184 Iter 2209, loss: 0.6127 Iter 2210, loss: 0.5722 Iter 2211, loss: 0.6116 Iter 2212, loss: 0.6234 Iter 2213, loss: 0.6071 Iter 2214, loss: 0.5997 Iter 2215, loss: 0.5780 Iter 2216, loss: 0.6344 Iter 2217, loss: 0.6372 Iter 2218, loss: 0.5993 Iter 2219, loss: 0.5952 Iter 2220, loss: 0.6086 Iter 2221, loss: 0.5764 Iter 2222, loss: 0.6062 Iter 2223, loss: 0.5978 Iter 2224, loss: 0.5974 Iter 2225, loss: 0.6089 Iter 2226, loss: 0.6064 Iter 2227, loss: 0.6683 Iter 2228, loss: 0.6109 Iter 2229, loss: 0.5632 Iter 2230, loss: 0.5923

Iter 2231, loss: 0.5583 Iter 2232, loss: 0.6102 Iter 2233, loss: 0.6079 Iter 2234, loss: 0.6375 Iter 2235, loss: 0.6010 Iter 2236, loss: 0.5952 Iter 2237, loss: 0.6336 Iter 2238, loss: 0.6158 Iter 2239, loss: 0.5939 Iter 2240, loss: 0.6135 Iter 2241, loss: 0.6482 Iter 2242, loss: 0.6172 Iter 2243, loss: 0.6028 Iter 2244, loss: 0.6596 Iter 2245, loss: 0.6373 Iter 2246, loss: 0.6143 Iter 2247, loss: 0.5777 Iter 2248, loss: 0.6163 Iter 2249, loss: 0.6075 Iter 2250, loss: 0.6233 Iter 2251, loss: 0.5677 Iter 2252, loss: 0.6217 Iter 2253, loss: 0.6566 Iter 2254, loss: 0.5910 Iter 2255, loss: 0.6111 Iter 2256, loss: 0.6616 Iter 2257, loss: 0.6073 Iter 2258, loss: 0.6183 Iter 2259, loss: 0.6212 Iter 2260, loss: 0.6159 Iter 2261, loss: 0.5633 Iter 2262, loss: 0.5735 Iter 2263, loss: 0.5634 Iter 2264, loss: 0.6138 Iter 2265, loss: 0.5964 Iter 2266, loss: 0.5966 Iter 2267, loss: 0.5464 Iter 2268, loss: 0.5928 Iter 2269, loss: 0.5912 Iter 2270, loss: 0.5824 Iter 2271, loss: 0.5838

Iter 2272, loss: 0.5831 Iter 2273, loss: 0.5962 Iter 2274, loss: 0.5589 Iter 2275, loss: 0.6090 Iter 2276, loss: 0.6046 Iter 2277, loss: 0.5934 Iter 2278, loss: 0.5620 Iter 2279, loss: 0.6419 Iter 2280, loss: 0.5948 Iter 2281, loss: 0.6439 Iter 2282, loss: 0.6005 Iter 2283, loss: 0.5956 Iter 2284, loss: 0.5818 Iter 2285, loss: 0.5924 Iter 2286, loss: 0.6065 Iter 2287, loss: 0.6166 Iter 2288, loss: 0.6036 Iter 2289, loss: 0.6051 Iter 2290, loss: 0.5926 Iter 2291, loss: 0.6408 Iter 2292, loss: 0.5840 Iter 2293, loss: 0.6136 Iter 2294, loss: 0.5971 Iter 2295, loss: 0.5882 Iter 2296, loss: 0.6440 Iter 2297, loss: 0.6475 Iter 2298, loss: 0.5986 Iter 2299, loss: 0.6354 Iter 2300, loss: 0.5845 Iter 2301, loss: 0.6363 Iter 2302, loss: 0.5802 Iter 2303, loss: 0.5860 Iter 2304, loss: 0.6033 Iter 2305, loss: 0.6059 Iter 2306, loss: 0.6260 Iter 2307, loss: 0.6046 Iter 2308, loss: 0.6398 Iter 2309, loss: 0.6304 Iter 2310, loss: 0.6076 Iter 2311, loss: 0.5773 Iter 2312, loss: 0.5753 Iter 2313, loss: 0.5894

Iter 2314, loss: 0.6038 Iter 2315, loss: 0.6319 Iter 2316, loss: 0.5971 Iter 2317, loss: 0.5815 Iter 2318, loss: 0.6205 Iter 2319, loss: 0.5999 Iter 2320, loss: 0.6370 Iter 2321, loss: 0.6429 Iter 2322, loss: 0.6273 Iter 2323, loss: 0.6320 Iter 2324, loss: 0.6513 Iter 2325, loss: 0.6224 Iter 2326, loss: 0.6196 Iter 2327, loss: 0.6062 Iter 2328, loss: 0.6107 Iter 2329, loss: 0.6365 Iter 2330, loss: 0.5799 Iter 2331, loss: 0.5870 Iter 2332, loss: 0.5974 Iter 2333, loss: 0.5749 Iter 2334, loss: 0.5795 Iter 2335, loss: 0.6127 Iter 2336, loss: 0.6027 Iter 2337, loss: 0.6425 Iter 2338, loss: 0.6417 Iter 2339, loss: 0.6261 Iter 2340, loss: 0.5766 Iter 2341, loss: 0.5426 Iter 2342, loss: 0.6396 Iter 2343, loss: 0.6460 Iter 2344, loss: 0.5965 Iter 2345, loss: 0.5754 Iter 2346, loss: 0.6025 Iter 2347, loss: 0.5814 Iter 2348, loss: 0.5709 Iter 2349, loss: 0.6305 Iter 2350, loss: 0.5925 Iter 2351, loss: 0.6115 Iter 2352, loss: 0.5915 Iter 2353, loss: 0.6203 Iter 2354, loss: 0.6476

Iter 2355, loss: 0.6108 Iter 2356, loss: 0.6158 Iter 2357, loss: 0.5818 Iter 2358, loss: 0.5976 Iter 2359, loss: 0.5851 Iter 2360, loss: 0.6005 Iter 2361, loss: 0.6279 Iter 2362, loss: 0.6025 Iter 2363, loss: 0.6008 Iter 2364, loss: 0.5647 Iter 2365, loss: 0.5637 Iter 2366, loss: 0.5684 Iter 2367, loss: 0.6227 Iter 2368, loss: 0.5956 Iter 2369, loss: 0.6139 Iter 2370, loss: 0.5911 Iter 2371, loss: 0.6346 Iter 2372, loss: 0.6232 Iter 2373, loss: 0.6012 Iter 2374, loss: 0.5749 Iter 2375, loss: 0.5991 Iter 2376, loss: 0.6335 Iter 2377, loss: 0.6435 Iter 2378, loss: 0.5973 Iter 2379, loss: 0.5849 Iter 2380, loss: 0.5680 Iter 2381, loss: 0.6363 Iter 2382, loss: 0.6001 Iter 2383, loss: 0.6167 Iter 2384, loss: 0.5662 Iter 2385, loss: 0.5786 Iter 2386, loss: 0.5989 Iter 2387, loss: 0.6254 Iter 2388, loss: 0.6169 Iter 2389, loss: 0.5644 Iter 2390, loss: 0.6124 Iter 2391, loss: 0.6290 Iter 2392, loss: 0.6265 Iter 2393, loss: 0.6048 Iter 2394, loss: 0.5782 Iter 2395, loss: 0.6339

Iter 2396, loss: 0.5813 Iter 2397, loss: 0.6165 Iter 2398, loss: 0.5834 Iter 2399, loss: 0.6122 Iter 2400, loss: 0.5964 Iter 2401, loss: 0.6379 Iter 2402, loss: 0.6335 Iter 2403, loss: 0.6289 Iter 2404, loss: 0.5807 Iter 2405, loss: 0.6664 Iter 2406, loss: 0.6437 Iter 2407, loss: 0.6262 Iter 2408, loss: 0.6083 Iter 2409, loss: 0.6717 Iter 2410, loss: 0.5958 Iter 2411, loss: 0.5987 Iter 2412, loss: 0.6124 Iter 2413, loss: 0.5686 Iter 2414, loss: 0.6504 Iter 2415, loss: 0.6192 Iter 2416, loss: 0.6153 Iter 2417, loss: 0.6181 Iter 2418, loss: 0.6056 Iter 2419, loss: 0.6536 Iter 2420, loss: 0.5761 Iter 2421, loss: 0.5808 Iter 2422, loss: 0.5739 Iter 2423, loss: 0.6230 Iter 2424, loss: 0.6372 Iter 2425, loss: 0.6323 Iter 2426, loss: 0.5428 Iter 2427, loss: 0.6274 Iter 2428, loss: 0.6181 Iter 2429, loss: 0.6154 Iter 2430, loss: 0.6193 Iter 2431, loss: 0.5937 Iter 2432, loss: 0.6407 Iter 2433, loss: 0.6209 Iter 2434, loss: 0.5966 Iter 2435, loss: 0.5612 Iter 2436, loss: 0.6091

Iter 2437, loss: 0.6115 Iter 2438, loss: 0.6008 Iter 2439, loss: 0.6040 Iter 2440, loss: 0.6240 Iter 2441, loss: 0.6041 Iter 2442, loss: 0.6341 Iter 2443, loss: 0.6181 Iter 2444, loss: 0.5847 Iter 2445, loss: 0.6015 Iter 2446, loss: 0.5382 Iter 2447, loss: 0.6151 Iter 2448, loss: 0.5532 Iter 2449, loss: 0.5970 Iter 2450, loss: 0.6104 Iter 2451, loss: 0.6326 Iter 2452, loss: 0.5973 Iter 2453, loss: 0.6054 Iter 2454, loss: 0.6265 Iter 2455, loss: 0.6032 Iter 2456, loss: 0.6014 Iter 2457, loss: 0.6233 Iter 2458, loss: 0.6581 Iter 2459, loss: 0.6405 Iter 2460, loss: 0.5824 Iter 2461, loss: 0.6410 Iter 2462, loss: 0.6287 Iter 2463, loss: 0.6095 Iter 2464, loss: 0.5863 Iter 2465, loss: 0.5931 Iter 2466, loss: 0.6260 Iter 2467, loss: 0.6101 Iter 2468, loss: 0.5832 Iter 2469, loss: 0.5945 Iter 2470, loss: 0.5940 Iter 2471, loss: 0.5771 Iter 2472, loss: 0.6231 Iter 2473, loss: 0.5664 Iter 2474, loss: 0.5965 Iter 2475, loss: 0.5586 Iter 2476, loss: 0.5794 Iter 2477, loss: 0.5562 Iter 2478, loss: 0.5575

Iter 2479, loss: 0.5950 Iter 2480, loss: 0.5569 Iter 2481, loss: 0.5591 Iter 2482, loss: 0.6250 Iter 2483, loss: 0.6202 Iter 2484, loss: 0.6478 Iter 2485, loss: 0.6122 Iter 2486, loss: 0.5913 Iter 2487, loss: 0.5785 Iter 2488, loss: 0.6275 Iter 2489, loss: 0.6018 Iter 2490, loss: 0.6151 Iter 2491, loss: 0.5630 Iter 2492, loss: 0.6158 Iter 2493, loss: 0.6000 Iter 2494, loss: 0.6038 Iter 2495, loss: 0.6182 Iter 2496, loss: 0.5723 Iter 2497, loss: 0.6250 Iter 2498, loss: 0.6197 Iter 2499, loss: 0.6017 Iter 2500, loss: 0.6230 Iter 2501, loss: 0.6176 Iter 2502, loss: 0.5964 Iter 2503, loss: 0.6592 Iter 2504, loss: 0.5831 Iter 2505, loss: 0.6231 Iter 2506, loss: 0.5272 Iter 2507, loss: 0.5873 Iter 2508, loss: 0.6123 Iter 2509, loss: 0.5847 Iter 2510, loss: 0.6207 Iter 2511, loss: 0.6045 Iter 2512, loss: 0.6118 Iter 2513, loss: 0.5670 Iter 2514, loss: 0.6561 Iter 2515, loss: 0.6125 Iter 2516, loss: 0.5983 Iter 2517, loss: 0.6511 Iter 2518, loss: 0.6362 Iter 2519, loss: 0.6489

Iter 2520, loss: 0.6295 Iter 2521, loss: 0.6479 Iter 2522, loss: 0.6044 Iter 2523, loss: 0.6562 Iter 2524, loss: 0.6101 Iter 2525, loss: 0.5991 Iter 2526, loss: 0.5485 Iter 2527, loss: 0.5975 Iter 2528, loss: 0.5467 Iter 2529, loss: 0.6143 Iter 2530, loss: 0.6362 Iter 2531, loss: 0.6105 Iter 2532, loss: 0.6128 Iter 2533, loss: 0.5800 Iter 2534, loss: 0.5923 Iter 2535, loss: 0.5811 Iter 2536, loss: 0.6360 Iter 2537, loss: 0.6113 Iter 2538, loss: 0.5959 Iter 2539, loss: 0.5958 Iter 2540, loss: 0.5780 Iter 2541, loss: 0.6234 Iter 2542, loss: 0.5984 Iter 2543, loss: 0.6219 Iter 2544, loss: 0.6296 Iter 2545, loss: 0.5694 Iter 2546, loss: 0.5753 Iter 2547, loss: 0.5919 Iter 2548, loss: 0.6183 Iter 2549, loss: 0.6021 Iter 2550, loss: 0.5969 Iter 2551, loss: 0.5691 Iter 2552, loss: 0.6615 Iter 2553, loss: 0.6199 Iter 2554, loss: 0.5604 Iter 2555, loss: 0.6222 Iter 2556, loss: 0.6081 Iter 2557, loss: 0.5826 Iter 2558, loss: 0.6147 Iter 2559, loss: 0.5945 Iter 2560, loss: 0.6119

Iter 2561, loss: 0.5963 Iter 2562, loss: 0.6174 Iter 2563, loss: 0.5951 Iter 2564, loss: 0.5594 Iter 2565, loss: 0.6487 Iter 2566, loss: 0.6522 Iter 2567, loss: 0.6132 Iter 2568, loss: 0.5674 Iter 2569, loss: 0.6088 Iter 2570, loss: 0.5443 Iter 2571, loss: 0.6294 Iter 2572, loss: 0.5845 Iter 2573, loss: 0.6194 Iter 2574, loss: 0.6169 Iter 2575, loss: 0.6213 Iter 2576, loss: 0.5831 Iter 2577, loss: 0.5715 Iter 2578, loss: 0.6038 Iter 2579, loss: 0.5790 Iter 2580, loss: 0.6266 Iter 2581, loss: 0.6306 Iter 2582, loss: 0.5718 Iter 2583, loss: 0.5928 Iter 2584, loss: 0.5773 Iter 2585, loss: 0.5884 Iter 2586, loss: 0.5892 Iter 2587, loss: 0.6443 Iter 2588, loss: 0.6623 Iter 2589, loss: 0.5936 Iter 2590, loss: 0.6445 Iter 2591, loss: 0.5752 Iter 2592, loss: 0.6095 Iter 2593, loss: 0.6181 Iter 2594, loss: 0.6239 Iter 2595, loss: 0.6007 Iter 2596, loss: 0.6402 Iter 2597, loss: 0.5475 Iter 2598, loss: 0.5771 Iter 2599, loss: 0.6074 Iter 2600, loss: 0.6117 Iter 2601, loss: 0.6565 Iter 2602, loss: 0.5943

Iter 2603, loss: 0.6070 Iter 2604, loss: 0.6317 Iter 2605, loss: 0.6246 Iter 2606, loss: 0.5732 Iter 2607, loss: 0.6513 Iter 2608, loss: 0.6135 Iter 2609, loss: 0.6014 Iter 2610, loss: 0.5754 Iter 2611, loss: 0.6036 Iter 2612, loss: 0.5858 Iter 2613, loss: 0.6027 Iter 2614, loss: 0.6014 Iter 2615, loss: 0.6107 Iter 2616, loss: 0.6048 Iter 2617, loss: 0.5920 Iter 2618, loss: 0.6152 Iter 2619, loss: 0.5844 Iter 2620, loss: 0.6130 Iter 2621, loss: 0.5892 Iter 2622, loss: 0.5874 Iter 2623, loss: 0.5921 Iter 2624, loss: 0.5693 Iter 2625, loss: 0.6078 Iter 2626, loss: 0.6243 Iter 2627, loss: 0.6218 Iter 2628, loss: 0.6245 Iter 2629, loss: 0.5905 Iter 2630, loss: 0.6064 Iter 2631, loss: 0.6293 Iter 2632, loss: 0.6165 Iter 2633, loss: 0.5955 Iter 2634, loss: 0.6144 Iter 2635, loss: 0.5878 Iter 2636, loss: 0.6052 Iter 2637, loss: 0.6000 Iter 2638, loss: 0.6024 Iter 2639, loss: 0.5752 Iter 2640, loss: 0.5975 Iter 2641, loss: 0.5925 Iter 2642, loss: 0.5695 Iter 2643, loss: 0.6280 Iter 2644, loss: 0.6355

Iter 2645, loss: 0.6493 Iter 2646, loss: 0.5914 Iter 2647, loss: 0.6327 Iter 2648, loss: 0.6038 Iter 2649, loss: 0.6125 Iter 2650, loss: 0.6078 Iter 2651, loss: 0.5878 Iter 2652, loss: 0.6030 Iter 2653, loss: 0.6143 Iter 2654, loss: 0.5842 Iter 2655, loss: 0.6532 Iter 2656, loss: 0.6628 Iter 2657, loss: 0.6175 Iter 2658, loss: 0.6240 Iter 2659, loss: 0.5773 Iter 2660, loss: 0.5678 Iter 2661, loss: 0.6049 Iter 2662, loss: 0.5761 Iter 2663, loss: 0.6086 Iter 2664, loss: 0.5797 Iter 2665, loss: 0.5879 Iter 2666, loss: 0.5903 Iter 2667, loss: 0.6081 Iter 2668, loss: 0.6109 Iter 2669, loss: 0.6248 Iter 2670, loss: 0.6110 Iter 2671, loss: 0.5859 Iter 2672, loss: 0.5891 Iter 2673, loss: 0.6057 Iter 2674, loss: 0.5788 Iter 2675, loss: 0.6340 Iter 2676, loss: 0.5867 Iter 2677, loss: 0.5526 Iter 2678, loss: 0.6225 Iter 2679, loss: 0.5850 Iter 2680, loss: 0.5890 Iter 2681, loss: 0.6215 Iter 2682, loss: 0.5382 Iter 2683, loss: 0.6079 Iter 2684, loss: 0.5932 Iter 2685, loss: 0.6138

Iter 2686, loss: 0.5858 Iter 2687, loss: 0.5465 Iter 2688, loss: 0.6240 Iter 2689, loss: 0.5935 Iter 2690, loss: 0.6266 Iter 2691, loss: 0.6146 Iter 2692, loss: 0.6005 Iter 2693, loss: 0.6376 Iter 2694, loss: 0.6208 Iter 2695, loss: 0.6198 Iter 2696, loss: 0.5947 Iter 2697, loss: 0.6310 Iter 2698, loss: 0.5880 Iter 2699, loss: 0.6444 Iter 2700, loss: 0.6015 Iter 2701, loss: 0.5855 Iter 2702, loss: 0.6072 Iter 2703, loss: 0.5744 Iter 2704, loss: 0.5673 Iter 2705, loss: 0.6278 Iter 2706, loss: 0.6209 Iter 2707, loss: 0.6138 Iter 2708, loss: 0.6384 Iter 2709, loss: 0.5908 Iter 2710, loss: 0.5881 Iter 2711, loss: 0.6054 Iter 2712, loss: 0.5707 Iter 2713, loss: 0.6206 Iter 2714, loss: 0.5906 Iter 2715, loss: 0.6397 Iter 2716, loss: 0.6009 Iter 2717, loss: 0.5979 Iter 2718, loss: 0.6363 Iter 2719, loss: 0.6235 Iter 2720, loss: 0.6036 Iter 2721, loss: 0.5919 Iter 2722, loss: 0.6390 Iter 2723, loss: 0.5746 Iter 2724, loss: 0.5702 Iter 2725, loss: 0.5819 Iter 2726, loss: 0.6000 Iter 2727, loss: 0.6187

Iter 2728, loss: 0.5876 Iter 2729, loss: 0.6247 Iter 2730, loss: 0.6320 Iter 2731, loss: 0.5688 Iter 2732, loss: 0.5621 Iter 2733, loss: 0.5900 Iter 2734, loss: 0.6402 Iter 2735, loss: 0.5955 Iter 2736, loss: 0.5806 Iter 2737, loss: 0.6077 Iter 2738, loss: 0.6003 Iter 2739, loss: 0.6365 Iter 2740, loss: 0.6281 Iter 2741, loss: 0.5827 Iter 2742, loss: 0.6014 Iter 2743, loss: 0.5906 Iter 2744, loss: 0.6355 Iter 2745, loss: 0.6106 Iter 2746, loss: 0.5560 Iter 2747, loss: 0.5903 Iter 2748, loss: 0.6255 Iter 2749, loss: 0.5445 Iter 2750, loss: 0.5769 Iter 2751, loss: 0.6498 Iter 2752, loss: 0.6069 Iter 2753, loss: 0.6364 Iter 2754, loss: 0.6303 Iter 2755, loss: 0.6525 Iter 2756, loss: 0.6027 Iter 2757, loss: 0.6154 Iter 2758, loss: 0.6182 Iter 2759, loss: 0.5882 Iter 2760, loss: 0.6295 Iter 2761, loss: 0.5867 Iter 2762, loss: 0.6424 Iter 2763, loss: 0.5699 Iter 2764, loss: 0.6637 Iter 2765, loss: 0.5995 Iter 2766, loss: 0.6715 Iter 2767, loss: 0.5939 Iter 2768, loss: 0.6271 Iter 2769, loss: 0.6365

Iter 2770, loss: 0.5958 Iter 2771, loss: 0.6324 Iter 2772, loss: 0.6597 Iter 2773, loss: 0.5868 Iter 2774, loss: 0.6092 Iter 2775, loss: 0.6001 Iter 2776, loss: 0.6114 Iter 2777, loss: 0.5864 Iter 2778, loss: 0.5748 Iter 2779, loss: 0.6089 Iter 2780, loss: 0.5684 Iter 2781, loss: 0.5941 Iter 2782, loss: 0.6397 Iter 2783, loss: 0.6401 Iter 2784, loss: 0.5777 Iter 2785, loss: 0.5799 Iter 2786, loss: 0.5986 Iter 2787, loss: 0.6765 Iter 2788, loss: 0.5576 Iter 2789, loss: 0.6334 Iter 2790, loss: 0.6018 Iter 2791, loss: 0.6083 Iter 2792, loss: 0.6074 Iter 2793, loss: 0.6016 Iter 2794, loss: 0.6024 Iter 2795, loss: 0.6069 Iter 2796, loss: 0.6131 Iter 2797, loss: 0.5784 Iter 2798, loss: 0.6668 Iter 2799, loss: 0.5979 Iter 2800, loss: 0.5776 Iter 2801, loss: 0.6023 Iter 2802, loss: 0.5656 Iter 2803, loss: 0.6082 Iter 2804, loss: 0.6263 Iter 2805, loss: 0.6292 Iter 2806, loss: 0.5781 Iter 2807, loss: 0.5479 Iter 2808, loss: 0.6112 Iter 2809, loss: 0.6001 Iter 2810, loss: 0.6199 Iter 2811, loss: 0.5741

Iter 2812, loss: 0.6081 Iter 2813, loss: 0.5692 Iter 2814, loss: 0.6315 Iter 2815, loss: 0.5967 Iter 2816, loss: 0.5942 Iter 2817, loss: 0.6257 Iter 2818, loss: 0.6084 Iter 2819, loss: 0.5984 Iter 2820, loss: 0.5940 Iter 2821, loss: 0.6185 Iter 2822, loss: 0.5969 Iter 2823, loss: 0.5960 Iter 2824, loss: 0.6288 Iter 2825, loss: 0.5930 Iter 2826, loss: 0.6391 Iter 2827, loss: 0.5794 Iter 2828, loss: 0.5607 Iter 2829, loss: 0.5609 Iter 2830, loss: 0.6341 Iter 2831, loss: 0.5982 Iter 2832, loss: 0.6294 Iter 2833, loss: 0.5917 Iter 2834, loss: 0.5649 Iter 2835, loss: 0.6176 Iter 2836, loss: 0.5477 Iter 2837, loss: 0.5887 Iter 2838, loss: 0.6044 Iter 2839, loss: 0.5994 Iter 2840, loss: 0.5474 Iter 2841, loss: 0.5802 Iter 2842, loss: 0.6032 Iter 2843, loss: 0.6538 Iter 2844, loss: 0.6434 Iter 2845, loss: 0.6298 Iter 2846, loss: 0.5920 Iter 2847, loss: 0.6359 Iter 2848, loss: 0.6307 Iter 2849, loss: 0.6054 Iter 2850, loss: 0.5769 Iter 2851, loss: 0.5790 Iter 2852, loss: 0.6170 Iter 2853, loss: 0.6009

Iter 2854, loss: 0.5872 Iter 2855, loss: 0.5799 Iter 2856, loss: 0.5797 Iter 2857, loss: 0.6386 Iter 2858, loss: 0.6031 Iter 2859, loss: 0.5911 Iter 2860, loss: 0.6198 Iter 2861, loss: 0.5952 Iter 2862, loss: 0.6453 Iter 2863, loss: 0.5866 Iter 2864, loss: 0.6320 Iter 2865, loss: 0.6404 Iter 2866, loss: 0.5848 Iter 2867, loss: 0.6000 Iter 2868, loss: 0.6212 Iter 2869, loss: 0.6208 Iter 2870, loss: 0.6481 Iter 2871, loss: 0.5784 Iter 2872, loss: 0.6380 Iter 2873, loss: 0.6053 Iter 2874, loss: 0.6019 Iter 2875, loss: 0.6196 Iter 2876, loss: 0.5694 Iter 2877, loss: 0.6075 Iter 2878, loss: 0.6130 Iter 2879, loss: 0.6068 Iter 2880, loss: 0.5999 Iter 2881, loss: 0.6286 Iter 2882, loss: 0.6037 Iter 2883, loss: 0.5888 Iter 2884, loss: 0.6130 Iter 2885, loss: 0.5504 Iter 2886, loss: 0.6048 Iter 2887, loss: 0.6351 Iter 2888, loss: 0.6382 Iter 2889, loss: 0.6328 Iter 2890, loss: 0.6195 Iter 2891, loss: 0.5999 Iter 2892, loss: 0.5626 Iter 2893, loss: 0.6291 Iter 2894, loss: 0.6097

Iter 2895, loss: 0.6446 Iter 2896, loss: 0.6153 Iter 2897, loss: 0.5927 Iter 2898, loss: 0.5729 Iter 2899, loss: 0.6128 Iter 2900, loss: 0.6124 Iter 2901, loss: 0.5950 Iter 2902, loss: 0.6164 Iter 2903, loss: 0.6096 Iter 2904, loss: 0.6020 Iter 2905, loss: 0.5784 Iter 2906, loss: 0.6109 Iter 2907, loss: 0.5951 Iter 2908, loss: 0.5874 Iter 2909, loss: 0.5994 Iter 2910, loss: 0.5923 Iter 2911, loss: 0.6086 Iter 2912, loss: 0.6281 Iter 2913, loss: 0.5911 Iter 2914, loss: 0.5909 Iter 2915, loss: 0.5978 Iter 2916, loss: 0.5782 Iter 2917, loss: 0.5870 Iter 2918, loss: 0.5914 Iter 2919, loss: 0.5937 Iter 2920, loss: 0.5655 Iter 2921, loss: 0.5847 Iter 2922, loss: 0.6067 Iter 2923, loss: 0.5906 Iter 2924, loss: 0.6012 Iter 2925, loss: 0.6436 Iter 2926, loss: 0.5539 Iter 2927, loss: 0.6008 Iter 2928, loss: 0.6106 Iter 2929, loss: 0.6021 Iter 2930, loss: 0.6544 Iter 2931, loss: 0.5490 Iter 2932, loss: 0.5992 Iter 2933, loss: 0.6260 Iter 2934, loss: 0.6133 Iter 2935, loss: 0.5800 Iter 2936, loss: 0.6106

Iter 2937, loss: 0.6086 Iter 2938, loss: 0.6324 Iter 2939, loss: 0.6010 Iter 2940, loss: 0.6133 Iter 2941, loss: 0.5750 Iter 2942, loss: 0.5872 Iter 2943, loss: 0.6390 Iter 2944, loss: 0.5765 Iter 2945, loss: 0.6087 Iter 2946, loss: 0.5816 Iter 2947, loss: 0.5978 Iter 2948, loss: 0.5888 Iter 2949, loss: 0.6175 Iter 2950, loss: 0.5818 Iter 2951, loss: 0.5737 Iter 2952, loss: 0.6159 Iter 2953, loss: 0.6118 Iter 2954, loss: 0.6188 Iter 2955, loss: 0.5920 Iter 2956, loss: 0.6224 Iter 2957, loss: 0.6127 Iter 2958, loss: 0.5558 Iter 2959, loss: 0.6113 Iter 2960, loss: 0.6089 Iter 2961, loss: 0.6016 Iter 2962, loss: 0.6474 Iter 2963, loss: 0.5721 Iter 2964, loss: 0.5760 Iter 2965, loss: 0.5906 Iter 2966, loss: 0.5623 Iter 2967, loss: 0.6130 Iter 2968, loss: 0.5126 Iter 2969, loss: 0.6078 Iter 2970, loss: 0.5612 Iter 2971, loss: 0.6347 Iter 2972, loss: 0.5878 Iter 2973, loss: 0.6359 Iter 2974, loss: 0.5856 Iter 2975, loss: 0.6201 Iter 2976, loss: 0.6085 Iter 2977, loss: 0.6048 Iter 2978, loss: 0.5873

Iter 2979, loss: 0.6274 Iter 2980, loss: 0.6034 Iter 2981, loss: 0.5996 Iter 2982, loss: 0.5918 Iter 2983, loss: 0.5820 Iter 2984, loss: 0.5642 Iter 2985, loss: 0.6077 Iter 2986, loss: 0.5857 Iter 2987, loss: 0.5836 Iter 2988, loss: 0.6272 Iter 2989, loss: 0.5714 Iter 2990, loss: 0.5312 Iter 2991, loss: 0.6004 Iter 2992, loss: 0.5761 Iter 2993, loss: 0.5738 Iter 2994, loss: 0.5685 Iter 2995, loss: 0.5509 Iter 2996, loss: 0.5868 Iter 2997, loss: 0.6262 Iter 2998, loss: 0.5612 Iter 2999, loss: 0.6126 Iter 3000, loss: 0.5710 Iter 3001, loss: 0.6125 Iter 3002, loss: 0.5761 Iter 3003, loss: 0.6698 Iter 3004, loss: 0.5711 Iter 3005, loss: 0.5752 Iter 3006, loss: 0.6431 Iter 3007, loss: 0.6402 Iter 3008, loss: 0.5924 Iter 3009, loss: 0.5881 Iter 3010, loss: 0.6052 Iter 3011, loss: 0.6561 Iter 3012, loss: 0.5963 Iter 3013, loss: 0.5878 Iter 3014, loss: 0.5772 Iter 3015, loss: 0.6548 Iter 3016, loss: 0.5881 Iter 3017, loss: 0.6195 Iter 3018, loss: 0.6147 Iter 3019, loss: 0.5394

Iter 3020, loss: 0.5828 Iter 3021, loss: 0.6169 Iter 3022, loss: 0.5548 Iter 3023, loss: 0.6470 Iter 3024, loss: 0.6084 Iter 3025, loss: 0.5795 Iter 3026, loss: 0.5967 Iter 3027, loss: 0.5860 Iter 3028, loss: 0.5912 Iter 3029, loss: 0.6275 Iter 3030, loss: 0.5649 Iter 3031, loss: 0.6003 Iter 3032, loss: 0.5635 Iter 3033, loss: 0.6041 Iter 3034, loss: 0.6244 Iter 3035, loss: 0.5739 Iter 3036, loss: 0.6019 Iter 3037, loss: 0.6500 Iter 3038, loss: 0.6083 Iter 3039, loss: 0.6055 Iter 3040, loss: 0.6149 Iter 3041, loss: 0.6062 Iter 3042, loss: 0.5760 Iter 3043, loss: 0.5532 Iter 3044, loss: 0.6279 Iter 3045, loss: 0.6042 Iter 3046, loss: 0.6205 Iter 3047, loss: 0.5930 Iter 3048, loss: 0.6062 Iter 3049, loss: 0.5554 Iter 3050, loss: 0.5675 Iter 3051, loss: 0.5580 Iter 3052, loss: 0.5929 Iter 3053, loss: 0.5893 Iter 3054, loss: 0.6451 Iter 3055, loss: 0.5824 Iter 3056, loss: 0.6146 Iter 3057, loss: 0.6001 Iter 3058, loss: 0.5830 Iter 3059, loss: 0.5999 Iter 3060, loss: 0.5938 Iter 3061, loss: 0.6137

Iter 3062, loss: 0.5988 Iter 3063, loss: 0.5873 Iter 3064, loss: 0.6419 Iter 3065, loss: 0.5892 Iter 3066, loss: 0.5898 Iter 3067, loss: 0.5719 Iter 3068, loss: 0.5374 Iter 3069, loss: 0.5914 Iter 3070, loss: 0.6322 Iter 3071, loss: 0.6349 Iter 3072, loss: 0.6423 Iter 3073, loss: 0.5793 Iter 3074, loss: 0.5662 Iter 3075, loss: 0.6004 Iter 3076, loss: 0.6299 Iter 3077, loss: 0.5664 Iter 3078, loss: 0.6080 Iter 3079, loss: 0.5594 Iter 3080, loss: 0.6139 Iter 3081, loss: 0.5943 Iter 3082, loss: 0.6109 Iter 3083, loss: 0.6035 Iter 3084, loss: 0.5775 Iter 3085, loss: 0.6173 Iter 3086, loss: 0.5958 Iter 3087, loss: 0.6209 Iter 3088, loss: 0.6424 Iter 3089, loss: 0.5891 Iter 3090, loss: 0.6075 Iter 3091, loss: 0.6027 Iter 3092, loss: 0.6065 Iter 3093, loss: 0.5746 Iter 3094, loss: 0.6222 Iter 3095, loss: 0.5569 Iter 3096, loss: 0.6175 Iter 3097, loss: 0.5932 Iter 3098, loss: 0.6028 Iter 3099, loss: 0.6020 Iter 3100, loss: 0.5761 Iter 3101, loss: 0.6049 Iter 3102, loss: 0.6074

Iter 3103, loss: 0.6285 Iter 3104, loss: 0.6350 Iter 3105, loss: 0.6394 Iter 3106, loss: 0.5721 Iter 3107, loss: 0.5866 Iter 3108, loss: 0.6008 Iter 3109, loss: 0.5985 Iter 3110, loss: 0.6081 Iter 3111, loss: 0.6048 Iter 3112, loss: 0.6015 Iter 3113, loss: 0.6163 Iter 3114, loss: 0.6079 Iter 3115, loss: 0.6498 Iter 3116, loss: 0.6510 Iter 3117, loss: 0.5729 Iter 3118, loss: 0.5779 Iter 3119, loss: 0.6584 Iter 3120, loss: 0.5713 Iter 3121, loss: 0.6588 Iter 3122, loss: 0.5604 Iter 3123, loss: 0.6561 Iter 3124, loss: 0.5626 Iter 3125, loss: 0.6091 Iter 3126, loss: 0.6193 Iter 3127, loss: 0.5736 Iter 3128, loss: 0.6042 Iter 3129, loss: 0.5964 Iter 3130, loss: 0.6541 Iter 3131, loss: 0.6036 Iter 3132, loss: 0.6274 Iter 3133, loss: 0.5856 Iter 3134, loss: 0.5852 Iter 3135, loss: 0.5849 Iter 3136, loss: 0.6021 Iter 3137, loss: 0.6435 Iter 3138, loss: 0.5614 Iter 3139, loss: 0.5922 Iter 3140, loss: 0.5783 Iter 3141, loss: 0.5884 Iter 3142, loss: 0.6563 Iter 3143, loss: 0.6074

Iter 3144, loss: 0.6117 Iter 3145, loss: 0.5699 Iter 3146, loss: 0.5877 Iter 3147, loss: 0.6220 Iter 3148, loss: 0.6492 Iter 3149, loss: 0.5696 Iter 3150, loss: 0.6426 Iter 3151, loss: 0.6131 Iter 3152, loss: 0.5962 Iter 3153, loss: 0.6216 Iter 3154, loss: 0.5646 Iter 3155, loss: 0.5991 Iter 3156, loss: 0.5948 Iter 3157, loss: 0.5953 Iter 3158, loss: 0.6095 Iter 3159, loss: 0.6138 Iter 3160, loss: 0.6203 Iter 3161, loss: 0.6259 Iter 3162, loss: 0.6048 Iter 3163, loss: 0.6207 Iter 3164, loss: 0.5474 Iter 3165, loss: 0.5960 Iter 3166, loss: 0.6244 Iter 3167, loss: 0.5241 Iter 3168, loss: 0.5637 Iter 3169, loss: 0.6676 Iter 3170, loss: 0.6200 Iter 3171, loss: 0.5829 Iter 3172, loss: 0.6425 Iter 3173, loss: 0.5766 Iter 3174, loss: 0.6154 Iter 3175, loss: 0.5712 Iter 3176, loss: 0.6269 Iter 3177, loss: 0.5579 Iter 3178, loss: 0.6004 Iter 3179, loss: 0.5581 Iter 3180, loss: 0.6024 Iter 3181, loss: 0.5782 Iter 3182, loss: 0.5956 Iter 3183, loss: 0.5994 Iter 3184, loss: 0.6249

Iter 3185, loss: 0.6087 Iter 3186, loss: 0.5901 Iter 3187, loss: 0.6051 Iter 3188, loss: 0.6245 Iter 3189, loss: 0.5937 Iter 3190, loss: 0.6561 Iter 3191, loss: 0.6131 Iter 3192, loss: 0.5957 Iter 3193, loss: 0.6234 Iter 3194, loss: 0.6095 Iter 3195, loss: 0.5997 Iter 3196, loss: 0.5916 Iter 3197, loss: 0.6034 Iter 3198, loss: 0.6058 Iter 3199, loss: 0.5657 Iter 3200, loss: 0.5528 Iter 3201, loss: 0.5361 Iter 3202, loss: 0.6388 Iter 3203, loss: 0.5627 Iter 3204, loss: 0.6053 Iter 3205, loss: 0.5710 Iter 3206, loss: 0.6128 Iter 3207, loss: 0.6271 Iter 3208, loss: 0.5951 Iter 3209, loss: 0.6319 Iter 3210, loss: 0.6207 Iter 3211, loss: 0.6108 Iter 3212, loss: 0.5559 Iter 3213, loss: 0.6687 Iter 3214, loss: 0.6642 Iter 3215, loss: 0.6419 Iter 3216, loss: 0.5879 Iter 3217, loss: 0.5949 Iter 3218, loss: 0.5975 Iter 3219, loss: 0.5878 Iter 3220, loss: 0.5623 Iter 3221, loss: 0.5761 Iter 3222, loss: 0.5937 Iter 3223, loss: 0.5690 Iter 3224, loss: 0.5511 Iter 3225, loss: 0.6399

Iter 3226, loss: 0.5791 Iter 3227, loss: 0.6206 Iter 3228, loss: 0.5638 Iter 3229, loss: 0.5824 Iter 3230, loss: 0.5900 Iter 3231, loss: 0.5858 Iter 3232, loss: 0.5670 Iter 3233, loss: 0.6071 Iter 3234, loss: 0.5816 Iter 3235, loss: 0.5966 Iter 3236, loss: 0.6337 Iter 3237, loss: 0.5953 Iter 3238, loss: 0.6110 Iter 3239, loss: 0.5892 Iter 3240, loss: 0.6035 Iter 3241, loss: 0.6381 Iter 3242, loss: 0.5932 Iter 3243, loss: 0.6402 Iter 3244, loss: 0.5696 Iter 3245, loss: 0.6259 Iter 3246, loss: 0.5742 Iter 3247, loss: 0.6575 Iter 3248, loss: 0.6279 Iter 3249, loss: 0.5636 Iter 3250, loss: 0.6577 Iter 3251, loss: 0.6289 Iter 3252, loss: 0.5756 Iter 3253, loss: 0.6267 Iter 3254, loss: 0.6163 Iter 3255, loss: 0.6580 Iter 3256, loss: 0.6220 Iter 3257, loss: 0.5547 Iter 3258, loss: 0.5767 Iter 3259, loss: 0.6134 Iter 3260, loss: 0.6329 Iter 3261, loss: 0.5998 Iter 3262, loss: 0.6110 Iter 3263, loss: 0.6077 Iter 3264, loss: 0.6008 Iter 3265, loss: 0.5615 Iter 3266, loss: 0.5613 Iter 3267, loss: 0.5689

Iter 3268, loss: 0.5971 Iter 3269, loss: 0.5654 Iter 3270, loss: 0.6114 Iter 3271, loss: 0.5844 Iter 3272, loss: 0.5786 Iter 3273, loss: 0.6204 Iter 3274, loss: 0.6016 Iter 3275, loss: 0.6283 Iter 3276, loss: 0.5880 Iter 3277, loss: 0.6276 Iter 3278, loss: 0.6590 Iter 3279, loss: 0.5205 Iter 3280, loss: 0.6528 Iter 3281, loss: 0.6088 Iter 3282, loss: 0.5570 Iter 3283, loss: 0.5918 Iter 3284, loss: 0.5970 Iter 3285, loss: 0.6451 Iter 3286, loss: 0.5778 Iter 3287, loss: 0.6399 Iter 3288, loss: 0.6034 Iter 3289, loss: 0.5829 Iter 3290, loss: 0.5691 Iter 3291, loss: 0.6182 Iter 3292, loss: 0.5558 Iter 3293, loss: 0.5759 Iter 3294, loss: 0.6158 Iter 3295, loss: 0.6010 Iter 3296, loss: 0.6312 Iter 3297, loss: 0.5820 Iter 3298, loss: 0.5718 Iter 3299, loss: 0.5861 Iter 3300, loss: 0.6251 Iter 3301, loss: 0.6084 Iter 3302, loss: 0.6081 Iter 3303, loss: 0.6262 Iter 3304, loss: 0.5803 Iter 3305, loss: 0.5937 Iter 3306, loss: 0.6111 Iter 3307, loss: 0.6253 Iter 3308, loss: 0.6412

Iter 3309, loss: 0.6295 Iter 3310, loss: 0.6328 Iter 3311, loss: 0.6231 Iter 3312, loss: 0.6293 Iter 3313, loss: 0.6028 Iter 3314, loss: 0.6024 Iter 3315, loss: 0.5880 Iter 3316, loss: 0.6123 Iter 3317, loss: 0.6356 Iter 3318, loss: 0.6112 Iter 3319, loss: 0.5758 Iter 3320, loss: 0.6239 Iter 3321, loss: 0.5511 Iter 3322, loss: 0.5817 Iter 3323, loss: 0.5962 Iter 3324, loss: 0.5988 Iter 3325, loss: 0.5960 Iter 3326, loss: 0.6140 Iter 3327, loss: 0.5663 Iter 3328, loss: 0.5925 Iter 3329, loss: 0.6062 Iter 3330, loss: 0.6082 Iter 3331, loss: 0.6208 Iter 3332, loss: 0.6355 Iter 3333, loss: 0.5656 Iter 3334, loss: 0.6209 Iter 3335, loss: 0.5777 Iter 3336, loss: 0.6112 Iter 3337, loss: 0.5892 Iter 3338, loss: 0.5952 Iter 3339, loss: 0.6021 Iter 3340, loss: 0.5646 Iter 3341, loss: 0.5836 Iter 3342, loss: 0.6153 Iter 3343, loss: 0.5991 Iter 3344, loss: 0.5569 Iter 3345, loss: 0.5812 Iter 3346, loss: 0.5971 Iter 3347, loss: 0.6169 Iter 3348, loss: 0.5793 Iter 3349, loss: 0.6370

Iter 3350, loss: 0.5533 Iter 3351, loss: 0.5695 Iter 3352, loss: 0.6273 Iter 3353, loss: 0.6151 Iter 3354, loss: 0.6371 Iter 3355, loss: 0.6278 Iter 3356, loss: 0.5938 Iter 3357, loss: 0.5502 Iter 3358, loss: 0.5723 Iter 3359, loss: 0.5678 Iter 3360, loss: 0.5867 Iter 3361, loss: 0.5947 Iter 3362, loss: 0.5588 Iter 3363, loss: 0.5851 Iter 3364, loss: 0.5787 Iter 3365, loss: 0.6010 Iter 3366, loss: 0.5783 Iter 3367, loss: 0.5760 Iter 3368, loss: 0.6120 Iter 3369, loss: 0.6131 Iter 3370, loss: 0.5680 Iter 3371, loss: 0.6076 Iter 3372, loss: 0.6583 Iter 3373, loss: 0.5651 Iter 3374, loss: 0.6331 Iter 3375, loss: 0.5622 Iter 3376, loss: 0.5983 Iter 3377, loss: 0.6075 Iter 3378, loss: 0.6222 Iter 3379, loss: 0.5784 Iter 3380, loss: 0.5771 Iter 3381, loss: 0.5866 Iter 3382, loss: 0.6233 Iter 3383, loss: 0.6390 Iter 3384, loss: 0.5422 Iter 3385, loss: 0.5687 Iter 3386, loss: 0.5979 Iter 3387, loss: 0.6205 Iter 3388, loss: 0.5831 Iter 3389, loss: 0.5546 Iter 3390, loss: 0.6338 Iter 3391, loss: 0.5582

Iter 3392, loss: 0.5935 Iter 3393, loss: 0.5844 Iter 3394, loss: 0.6476 Iter 3395, loss: 0.6153 Iter 3396, loss: 0.5656 Iter 3397, loss: 0.5988 Iter 3398, loss: 0.5640 Iter 3399, loss: 0.5858 Iter 3400, loss: 0.6239 Iter 3401, loss: 0.5506 Iter 3402, loss: 0.6033 Iter 3403, loss: 0.5964 Iter 3404, loss: 0.5851 Iter 3405, loss: 0.5760 Iter 3406, loss: 0.5793 Iter 3407, loss: 0.5924 Iter 3408, loss: 0.6227 Iter 3409, loss: 0.6182 Iter 3410, loss: 0.5851 Iter 3411, loss: 0.5841 Iter 3412, loss: 0.6009 Iter 3413, loss: 0.5604 Iter 3414, loss: 0.6111 Iter 3415, loss: 0.6029 Iter 3416, loss: 0.5674 Iter 3417, loss: 0.6004 Iter 3418, loss: 0.6083 Iter 3419, loss: 0.5353 Iter 3420, loss: 0.5900 Iter 3421, loss: 0.6503 Iter 3422, loss: 0.5901 Iter 3423, loss: 0.5616 Iter 3424, loss: 0.5955 Iter 3425, loss: 0.6231 Iter 3426, loss: 0.6066 Iter 3427, loss: 0.5801 Iter 3428, loss: 0.5993 Iter 3429, loss: 0.5781 Iter 3430, loss: 0.6593 Iter 3431, loss: 0.5502 Iter 3432, loss: 0.5873

Iter 3433, loss: 0.5955 Iter 3434, loss: 0.6432 Iter 3435, loss: 0.6544 Iter 3436, loss: 0.6146 Iter 3437, loss: 0.6155 Iter 3438, loss: 0.6222 Iter 3439, loss: 0.5693 Iter 3440, loss: 0.6084 Iter 3441, loss: 0.5744 Iter 3442, loss: 0.5703 Iter 3443, loss: 0.5746 Iter 3444, loss: 0.5483 Iter 3445, loss: 0.5633 Iter 3446, loss: 0.6434 Iter 3447, loss: 0.5937 Iter 3448, loss: 0.5487 Iter 3449, loss: 0.6019 Iter 3450, loss: 0.6192 Iter 3451, loss: 0.5805 Iter 3452, loss: 0.5595 Iter 3453, loss: 0.5908 Iter 3454, loss: 0.6216 Iter 3455, loss: 0.6704 Iter 3456, loss: 0.5744 Iter 3457, loss: 0.6132 Iter 3458, loss: 0.5841 Iter 3459, loss: 0.6459 Iter 3460, loss: 0.5965 Iter 3461, loss: 0.5943 Iter 3462, loss: 0.5822 Iter 3463, loss: 0.6179 Iter 3464, loss: 0.5342 Iter 3465, loss: 0.5701 Iter 3466, loss: 0.5924 Iter 3467, loss: 0.5897 Iter 3468, loss: 0.5700 Iter 3469, loss: 0.6056 Iter 3470, loss: 0.5718 Iter 3471, loss: 0.6296 Iter 3472, loss: 0.6257 Iter 3473, loss: 0.5792 Iter 3474, loss: 0.5887

Iter 3475, loss: 0.5481 Iter 3476, loss: 0.5758 Iter 3477, loss: 0.5905 Iter 3478, loss: 0.5682 Iter 3479, loss: 0.5896 Iter 3480, loss: 0.5975 Iter 3481, loss: 0.6332 Iter 3482, loss: 0.5947 Iter 3483, loss: 0.5343 Iter 3484, loss: 0.6210 Iter 3485, loss: 0.5778 Iter 3486, loss: 0.5940 Iter 3487, loss: 0.6021 Iter 3488, loss: 0.5784 Iter 3489, loss: 0.6046 Iter 3490, loss: 0.5922 Iter 3491, loss: 0.6102 Iter 3492, loss: 0.5923 Iter 3493, loss: 0.5776 Iter 3494, loss: 0.6066 Iter 3495, loss: 0.5806 Iter 3496, loss: 0.5754 Iter 3497, loss: 0.6161 Iter 3498, loss: 0.6051 Iter 3499, loss: 0.5454 Iter 3500, loss: 0.6035 Iter 3501, loss: 0.5827 Iter 3502, loss: 0.6038 Iter 3503, loss: 0.5748 Iter 3504, loss: 0.5755 Iter 3505, loss: 0.5732 Iter 3506, loss: 0.5962 Iter 3507, loss: 0.6133 Iter 3508, loss: 0.5815 Iter 3509, loss: 0.5934 Iter 3510, loss: 0.6138 Iter 3511, loss: 0.6024 Iter 3512, loss: 0.5882 Iter 3513, loss: 0.5853 Iter 3514, loss: 0.5777 Iter 3515, loss: 0.5489

Iter 3516, loss: 0.6134 Iter 3517, loss: 0.5596 Iter 3518, loss: 0.5813 Iter 3519, loss: 0.6198 Iter 3520, loss: 0.5818 Iter 3521, loss: 0.6030 Iter 3522, loss: 0.5742 Iter 3523, loss: 0.5951 Iter 3524, loss: 0.5602 Iter 3525, loss: 0.6184 Iter 3526, loss: 0.6100 Iter 3527, loss: 0.6140 Iter 3528, loss: 0.5828 Iter 3529, loss: 0.5974 Iter 3530, loss: 0.5988 Iter 3531, loss: 0.5824 Iter 3532, loss: 0.6061 Iter 3533, loss: 0.5337 Iter 3534, loss: 0.5628 Iter 3535, loss: 0.6106 Iter 3536, loss: 0.6475 Iter 3537, loss: 0.5854 Iter 3538, loss: 0.5720 Iter 3539, loss: 0.5802 Iter 3540, loss: 0.5725 Iter 3541, loss: 0.6192 Iter 3542, loss: 0.5614 Iter 3543, loss: 0.5827 Iter 3544, loss: 0.5414 Iter 3545, loss: 0.5625 Iter 3546, loss: 0.6127 Iter 3547, loss: 0.5843 Iter 3548, loss: 0.6196 Iter 3549, loss: 0.5948 Iter 3550, loss: 0.5799 Iter 3551, loss: 0.5939 Iter 3552, loss: 0.6194 Iter 3553, loss: 0.5811 Iter 3554, loss: 0.6432 Iter 3555, loss: 0.5902 Iter 3556, loss: 0.6157 Iter 3557, loss: 0.5701

Iter 3558, loss: 0.5934 Iter 3559, loss: 0.5903 Iter 3560, loss: 0.6211 Iter 3561, loss: 0.5931 Iter 3562, loss: 0.5994 Iter 3563, loss: 0.6230 Iter 3564, loss: 0.6103 Iter 3565, loss: 0.5837 Iter 3566, loss: 0.5850 Iter 3567, loss: 0.6421 Iter 3568, loss: 0.5989 Iter 3569, loss: 0.5757 Iter 3570, loss: 0.6245 Iter 3571, loss: 0.5795 Iter 3572, loss: 0.5934 Iter 3573, loss: 0.5989 Iter 3574, loss: 0.5937 Iter 3575, loss: 0.5997 Iter 3576, loss: 0.6273 Iter 3577, loss: 0.5767 Iter 3578, loss: 0.5848 Iter 3579, loss: 0.5932 Iter 3580, loss: 0.5953 Iter 3581, loss: 0.6035 Iter 3582, loss: 0.5614 Iter 3583, loss: 0.5830 Iter 3584, loss: 0.6032 Iter 3585, loss: 0.5606 Iter 3586, loss: 0.6075 Iter 3587, loss: 0.5944 Iter 3588, loss: 0.6130 Iter 3589, loss: 0.5640 Iter 3590, loss: 0.6337 Iter 3591, loss: 0.6241 Iter 3592, loss: 0.6161 Iter 3593, loss: 0.6125 Iter 3594, loss: 0.5528 Iter 3595, loss: 0.5946 Iter 3596, loss: 0.6509 Iter 3597, loss: 0.6596 Iter 3598, loss: 0.5499 Iter 3599, loss: 0.6091

Iter 3600, loss: 0.6255 Iter 3601, loss: 0.5802 Iter 3602, loss: 0.5849 Iter 3603, loss: 0.6009 Iter 3604, loss: 0.5589 Iter 3605, loss: 0.6345 Iter 3606, loss: 0.5862 Iter 3607, loss: 0.5769 Iter 3608, loss: 0.5504 Iter 3609, loss: 0.5882 Iter 3610, loss: 0.6347 Iter 3611, loss: 0.6199 Iter 3612, loss: 0.6155 Iter 3613, loss: 0.6190 Iter 3614, loss: 0.5875 Iter 3615, loss: 0.5930 Iter 3616, loss: 0.5878 Iter 3617, loss: 0.5460 Iter 3618, loss: 0.5636 Iter 3619, loss: 0.6039 Iter 3620, loss: 0.5741 Iter 3621, loss: 0.5478 Iter 3622, loss: 0.6247 Iter 3623, loss: 0.5992 Iter 3624, loss: 0.6060 Iter 3625, loss: 0.6035 Iter 3626, loss: 0.5661 Iter 3627, loss: 0.5844 Iter 3628, loss: 0.5996 Iter 3629, loss: 0.5890 Iter 3630, loss: 0.6085 Iter 3631, loss: 0.6231 Iter 3632, loss: 0.6389 Iter 3633, loss: 0.6054 Iter 3634, loss: 0.5958 Iter 3635, loss: 0.6505 Iter 3636, loss: 0.6195 Iter 3637, loss: 0.5869 Iter 3638, loss: 0.6293 Iter 3639, loss: 0.5949 Iter 3640, loss: 0.6279

Iter 3641, loss: 0.5534 Iter 3642, loss: 0.5616 Iter 3643, loss: 0.5701 Iter 3644, loss: 0.5988 Iter 3645, loss: 0.6235 Iter 3646, loss: 0.6000 Iter 3647, loss: 0.5877 Iter 3648, loss: 0.5919 Iter 3649, loss: 0.5839 Iter 3650, loss: 0.5862 Iter 3651, loss: 0.6132 Iter 3652, loss: 0.6223 Iter 3653, loss: 0.5857 Iter 3654, loss: 0.5686 Iter 3655, loss: 0.5583 Iter 3656, loss: 0.6247 Iter 3657, loss: 0.5922 Iter 3658, loss: 0.5851 Iter 3659, loss: 0.5881 Iter 3660, loss: 0.5834 Iter 3661, loss: 0.6170 Iter 3662, loss: 0.5805 Iter 3663, loss: 0.5516 Iter 3664, loss: 0.6042 Iter 3665, loss: 0.6226 Iter 3666, loss: 0.5784 Iter 3667, loss: 0.6318 Iter 3668, loss: 0.5992 Iter 3669, loss: 0.5645 Iter 3670, loss: 0.6045 Iter 3671, loss: 0.5607 Iter 3672, loss: 0.6015 Iter 3673, loss: 0.5615 Iter 3674, loss: 0.6282 Iter 3675, loss: 0.5962 Iter 3676, loss: 0.5737 Iter 3677, loss: 0.6067 Iter 3678, loss: 0.5712 Iter 3679, loss: 0.5427 Iter 3680, loss: 0.5999 Iter 3681, loss: 0.6082 Iter 3682, loss: 0.5495

Iter 3683, loss: 0.5973 Iter 3684, loss: 0.6132 Iter 3685, loss: 0.5623 Iter 3686, loss: 0.5748 Iter 3687, loss: 0.6274 Iter 3688, loss: 0.5746 Iter 3689, loss: 0.5215 Iter 3690, loss: 0.6154 Iter 3691, loss: 0.5984 Iter 3692, loss: 0.6389 Iter 3693, loss: 0.5980 Iter 3694, loss: 0.5894 Iter 3695, loss: 0.5921 Iter 3696, loss: 0.5889 Iter 3697, loss: 0.5779 Iter 3698, loss: 0.5795 Iter 3699, loss: 0.5728 Iter 3700, loss: 0.5772 Iter 3701, loss: 0.6028 Iter 3702, loss: 0.5891 Iter 3703, loss: 0.5571 Iter 3704, loss: 0.5447 Iter 3705, loss: 0.6327 Iter 3706, loss: 0.6091 Iter 3707, loss: 0.5929 Iter 3708, loss: 0.5410 Iter 3709, loss: 0.5813 Iter 3710, loss: 0.5472 Iter 3711, loss: 0.5909 Iter 3712, loss: 0.5848 Iter 3713, loss: 0.5226 Iter 3714, loss: 0.6160 Iter 3715, loss: 0.6178 Iter 3716, loss: 0.6006 Iter 3717, loss: 0.6305 Iter 3718, loss: 0.5946 Iter 3719, loss: 0.6014 Iter 3720, loss: 0.5685 Iter 3721, loss: 0.5521 Iter 3722, loss: 0.6208 Iter 3723, loss: 0.6413 Iter 3724, loss: 0.6320

Iter 3725, loss: 0.5984 Iter 3726, loss: 0.5796 Iter 3727, loss: 0.6112 Iter 3728, loss: 0.5257 Iter 3729, loss: 0.6148 Iter 3730, loss: 0.5704 Iter 3731, loss: 0.5677 Iter 3732, loss: 0.6113 Iter 3733, loss: 0.5895 Iter 3734, loss: 0.5891 Iter 3735, loss: 0.5777 Iter 3736, loss: 0.5788 Iter 3737, loss: 0.6276 Iter 3738, loss: 0.6225 Iter 3739, loss: 0.5652 Iter 3740, loss: 0.6087 Iter 3741, loss: 0.5763 Iter 3742, loss: 0.5937 Iter 3743, loss: 0.6101 Iter 3744, loss: 0.6070 Iter 3745, loss: 0.5681 Iter 3746, loss: 0.6043 Iter 3747, loss: 0.5906 Iter 3748, loss: 0.5976 Iter 3749, loss: 0.6023 Iter 3750, loss: 0.5936 Iter 3751, loss: 0.5934 Iter 3752, loss: 0.6166 Iter 3753, loss: 0.6038 Iter 3754, loss: 0.6112 Iter 3755, loss: 0.5795 Iter 3756, loss: 0.6072 Iter 3757, loss: 0.5966 Iter 3758, loss: 0.5794 Iter 3759, loss: 0.6115 Iter 3760, loss: 0.6410 Iter 3761, loss: 0.6290 Iter 3762, loss: 0.6253 Iter 3763, loss: 0.5667 Iter 3764, loss: 0.5908 Iter 3765, loss: 0.6040

Iter 3766, loss: 0.6236 Iter 3767, loss: 0.6104 Iter 3768, loss: 0.6103 Iter 3769, loss: 0.6081 Iter 3770, loss: 0.5771 Iter 3771, loss: 0.6172 Iter 3772, loss: 0.5435 Iter 3773, loss: 0.5927 Iter 3774, loss: 0.5477 Iter 3775, loss: 0.5623 Iter 3776, loss: 0.5705 Iter 3777, loss: 0.6056 Iter 3778, loss: 0.5896 Iter 3779, loss: 0.5950 Iter 3780, loss: 0.5691 Iter 3781, loss: 0.5981 Iter 3782, loss: 0.5823 Iter 3783, loss: 0.6052 Iter 3784, loss: 0.5604 Iter 3785, loss: 0.6011 Iter 3786, loss: 0.6129 Iter 3787, loss: 0.5996 Iter 3788, loss: 0.6157 Iter 3789, loss: 0.5734 Iter 3790, loss: 0.5911 Iter 3791, loss: 0.5635 Iter 3792, loss: 0.5934 Iter 3793, loss: 0.6090 Iter 3794, loss: 0.5937 Iter 3795, loss: 0.5403 Iter 3796, loss: 0.5608 Iter 3797, loss: 0.6072 Iter 3798, loss: 0.5923 Iter 3799, loss: 0.5587 Iter 3800, loss: 0.5887 Iter 3801, loss: 0.6530 Iter 3802, loss: 0.5297 Iter 3803, loss: 0.5904 Iter 3804, loss: 0.5673 Iter 3805, loss: 0.6168 Iter 3806, loss: 0.5845 Iter 3807, loss: 0.6058

Iter 3808, loss: 0.6008 Iter 3809, loss: 0.6064 Iter 3810, loss: 0.5681 Iter 3811, loss: 0.5462 Iter 3812, loss: 0.5979 Iter 3813, loss: 0.6310 Iter 3814, loss: 0.5710 Iter 3815, loss: 0.5790 Iter 3816, loss: 0.5834 Iter 3817, loss: 0.5877 Iter 3818, loss: 0.5886 Iter 3819, loss: 0.5977 Iter 3820, loss: 0.6084 Iter 3821, loss: 0.5970 Iter 3822, loss: 0.5677 Iter 3823, loss: 0.5961 Iter 3824, loss: 0.6027 Iter 3825, loss: 0.6249 Iter 3826, loss: 0.5981 Iter 3827, loss: 0.5590 Iter 3828, loss: 0.6511 Iter 3829, loss: 0.6222 Iter 3830, loss: 0.6211 Iter 3831, loss: 0.5878 Iter 3832, loss: 0.6009 Iter 3833, loss: 0.5839 Iter 3834, loss: 0.6000 Iter 3835, loss: 0.6352 Iter 3836, loss: 0.5858 Iter 3837, loss: 0.5717 Iter 3838, loss: 0.5958 Iter 3839, loss: 0.6263 Iter 3840, loss: 0.6117 Iter 3841, loss: 0.6091 Iter 3842, loss: 0.5651 Iter 3843, loss: 0.6077 Iter 3844, loss: 0.6098 Iter 3845, loss: 0.5909 Iter 3846, loss: 0.5867 Iter 3847, loss: 0.6146 Iter 3848, loss: 0.5935 Iter 3849, loss: 0.5844

Iter 3850, loss: 0.5981 Iter 3851, loss: 0.6029 Iter 3852, loss: 0.5965 Iter 3853, loss: 0.5832 Iter 3854, loss: 0.5773 Iter 3855, loss: 0.6030 Iter 3856, loss: 0.5654 Iter 3857, loss: 0.6148 Iter 3858, loss: 0.5866 Iter 3859, loss: 0.6294 Iter 3860, loss: 0.6160 Iter 3861, loss: 0.5987 Iter 3862, loss: 0.6018 Iter 3863, loss: 0.6056 Iter 3864, loss: 0.5982 Iter 3865, loss: 0.5925 Iter 3866, loss: 0.5957 Iter 3867, loss: 0.5866 Iter 3868, loss: 0.6094 Iter 3869, loss: 0.6277 Iter 3870, loss: 0.5987 Iter 3871, loss: 0.6195 Iter 3872, loss: 0.6314 Iter 3873, loss: 0.6296 Iter 3874, loss: 0.5787 Iter 3875, loss: 0.6306 Iter 3876, loss: 0.5744 Iter 3877, loss: 0.5608 Iter 3878, loss: 0.5971 Iter 3879, loss: 0.5573 Iter 3880, loss: 0.5743 Iter 3881, loss: 0.6410 Iter 3882, loss: 0.6003 Iter 3883, loss: 0.5965 Iter 3884, loss: 0.5977 Iter 3885, loss: 0.6125 Iter 3886, loss: 0.6166 Iter 3887, loss: 0.5735 Iter 3888, loss: 0.5943 Iter 3889, loss: 0.6077 Iter 3890, loss: 0.5790 Iter 3891, loss: 0.5764

Iter 3892, loss: 0.6106 Iter 3893, loss: 0.6043 Iter 3894, loss: 0.6009 Iter 3895, loss: 0.5912 Iter 3896, loss: 0.5892 Iter 3897, loss: 0.6091 Iter 3898, loss: 0.5731 Iter 3899, loss: 0.5986 Iter 3900, loss: 0.5767 Iter 3901, loss: 0.6173 Iter 3902, loss: 0.5735 Iter 3903, loss: 0.5821 Iter 3904, loss: 0.5785 Iter 3905, loss: 0.5937 Iter 3906, loss: 0.5733 Iter 3907, loss: 0.5886 Iter 3908, loss: 0.6117 Iter 3909, loss: 0.5940 Iter 3910, loss: 0.5824 Iter 3911, loss: 0.5839 Iter 3912, loss: 0.6232 Iter 3913, loss: 0.5805 Iter 3914, loss: 0.5914 Iter 3915, loss: 0.5756 Iter 3916, loss: 0.5744 Iter 3917, loss: 0.6140 Iter 3918, loss: 0.6007 Iter 3919, loss: 0.5912 Iter 3920, loss: 0.5715 Iter 3921, loss: 0.6281 Iter 3922, loss: 0.5543 Iter 3923, loss: 0.6402 Iter 3924, loss: 0.6264 Iter 3925, loss: 0.5819 Iter 3926, loss: 0.5734 Iter 3927, loss: 0.6124 Iter 3928, loss: 0.6412 Iter 3929, loss: 0.5782 Iter 3930, loss: 0.5867 Iter 3931, loss: 0.5933 Iter 3932, loss: 0.5874

Iter 3933, loss: 0.6133 Iter 3934, loss: 0.6059 Iter 3935, loss: 0.5277 Iter 3936, loss: 0.6002 Iter 3937, loss: 0.6079 Iter 3938, loss: 0.6092 Iter 3939, loss: 0.5765 Iter 3940, loss: 0.5660 Iter 3941, loss: 0.5930 Iter 3942, loss: 0.6232 Iter 3943, loss: 0.6276 Iter 3944, loss: 0.5719 Iter 3945, loss: 0.5864 Iter 3946, loss: 0.5856 Iter 3947, loss: 0.6228 Iter 3948, loss: 0.5817 Iter 3949, loss: 0.6022 Iter 3950, loss: 0.5643 Iter 3951, loss: 0.5874 Iter 3952, loss: 0.6288 Iter 3953, loss: 0.6037 Iter 3954, loss: 0.5832 Iter 3955, loss: 0.5863 Iter 3956, loss: 0.6446 Iter 3957, loss: 0.5828 Iter 3958, loss: 0.5975 Iter 3959, loss: 0.5824 Iter 3960, loss: 0.6180 Iter 3961, loss: 0.5927 Iter 3962, loss: 0.5716 Iter 3963, loss: 0.6296 Iter 3964, loss: 0.5698 Iter 3965, loss: 0.6119 Iter 3966, loss: 0.6097 Iter 3967, loss: 0.5992 Iter 3968, loss: 0.5874 Iter 3969, loss: 0.6181 Iter 3970, loss: 0.5827 Iter 3971, loss: 0.5951 Iter 3972, loss: 0.5964 Iter 3973, loss: 0.5802

Iter 3974, loss: 0.6386 Iter 3975, loss: 0.5901 Iter 3976, loss: 0.5827 Iter 3977, loss: 0.5984 Iter 3978, loss: 0.5894 Iter 3979, loss: 0.6092 Iter 3980, loss: 0.5949 Iter 3981, loss: 0.6066 Iter 3982, loss: 0.5908 Iter 3983, loss: 0.6148 Iter 3984, loss: 0.6080 Iter 3985, loss: 0.6379 Iter 3986, loss: 0.6028 Iter 3987, loss: 0.5539 Iter 3988, loss: 0.5558 Iter 3989, loss: 0.5928 Iter 3990, loss: 0.5824 Iter 3991, loss: 0.5503 Iter 3992, loss: 0.6131 Iter 3993, loss: 0.6276 Iter 3994, loss: 0.5939 Iter 3995, loss: 0.5787 Iter 3996, loss: 0.5505 Iter 3997, loss: 0.5482 Iter 3998, loss: 0.5484 Iter 3999, loss: 0.5458 Iter 4000, loss: 0.6003 Iter 4001, loss: 0.6006 Iter 4002, loss: 0.6422 Iter 4003, loss: 0.5931 Iter 4004, loss: 0.6110 Iter 4005, loss: 0.5952 Iter 4006, loss: 0.6301 Iter 4007, loss: 0.5585 Iter 4008, loss: 0.6238 Iter 4009, loss: 0.5956 Iter 4010, loss: 0.5573 Iter 4011, loss: 0.5795 Iter 4012, loss: 0.6082 Iter 4013, loss: 0.5652 Iter 4014, loss: 0.6147 Iter 4015, loss: 0.5979

Iter 4016, loss: 0.5811 Iter 4017, loss: 0.5909 Iter 4018, loss: 0.5680 Iter 4019, loss: 0.5881 Iter 4020, loss: 0.6116 Iter 4021, loss: 0.5960 Iter 4022, loss: 0.5417 Iter 4023, loss: 0.6233 Iter 4024, loss: 0.5748 Iter 4025, loss: 0.6234 Iter 4026, loss: 0.6177 Iter 4027, loss: 0.6244 Iter 4028, loss: 0.5937 Iter 4029, loss: 0.5562 Iter 4030, loss: 0.5652 Iter 4031, loss: 0.5910 Iter 4032, loss: 0.5339 Iter 4033, loss: 0.6457 Iter 4034, loss: 0.6099 Iter 4035, loss: 0.5865 Iter 4036, loss: 0.6047 Iter 4037, loss: 0.5941 Iter 4038, loss: 0.5717 Iter 4039, loss: 0.6389 Iter 4040, loss: 0.5801 Iter 4041, loss: 0.6120 Iter 4042, loss: 0.5763 Iter 4043, loss: 0.6115 Iter 4044, loss: 0.5742 Iter 4045, loss: 0.6150 Iter 4046, loss: 0.5427 Iter 4047, loss: 0.5982 Iter 4048, loss: 0.6151 Iter 4049, loss: 0.5647 Iter 4050, loss: 0.5792 Iter 4051, loss: 0.5494 Iter 4052, loss: 0.5842 Iter 4053, loss: 0.6101 Iter 4054, loss: 0.6410 Iter 4055, loss: 0.5816 Iter 4056, loss: 0.5618 Iter 4057, loss: 0.5584

Iter 4058, loss: 0.6020 Iter 4059, loss: 0.6141 Iter 4060, loss: 0.5764 Iter 4061, loss: 0.5474 Iter 4062, loss: 0.5733 Iter 4063, loss: 0.5632 Iter 4064, loss: 0.5901 Iter 4065, loss: 0.5787 Iter 4066, loss: 0.6315 Iter 4067, loss: 0.6061 Iter 4068, loss: 0.5692 Iter 4069, loss: 0.6101 Iter 4070, loss: 0.5807 Iter 4071, loss: 0.5792 Iter 4072, loss: 0.6563 Iter 4073, loss: 0.5981 Iter 4074, loss: 0.6424 Iter 4075, loss: 0.6084 Iter 4076, loss: 0.6063 Iter 4077, loss: 0.6510 Iter 4078, loss: 0.5831 Iter 4079, loss: 0.5366 Iter 4080, loss: 0.6278 Iter 4081, loss: 0.5557 Iter 4082, loss: 0.5855 Iter 4083, loss: 0.5789 Iter 4084, loss: 0.6284 Iter 4085, loss: 0.5994 Iter 4086, loss: 0.6104 Iter 4087, loss: 0.6081 Iter 4088, loss: 0.5675 Iter 4089, loss: 0.6055 Iter 4090, loss: 0.5842 Iter 4091, loss: 0.6000 Iter 4092, loss: 0.5812 Iter 4093, loss: 0.6288 Iter 4094, loss: 0.6096 Iter 4095, loss: 0.6094 Iter 4096, loss: 0.5972 Iter 4097, loss: 0.5553 Iter 4098, loss: 0.5941 Iter 4099, loss: 0.6239

Iter 4100, loss: 0.5852 Iter 4101, loss: 0.5741 Iter 4102, loss: 0.5836 Iter 4103, loss: 0.6057 Iter 4104, loss: 0.5739 Iter 4105, loss: 0.5643 Iter 4106, loss: 0.6326 Iter 4107, loss: 0.6089 Iter 4108, loss: 0.6314 Iter 4109, loss: 0.6111 Iter 4110, loss: 0.6080 Iter 4111, loss: 0.5782 Iter 4112, loss: 0.6241 Iter 4113, loss: 0.6016 Iter 4114, loss: 0.5808 Iter 4115, loss: 0.5552 Iter 4116, loss: 0.5622 Iter 4117, loss: 0.5895 Iter 4118, loss: 0.6004 Iter 4119, loss: 0.5831 Iter 4120, loss: 0.5671 Iter 4121, loss: 0.5786 Iter 4122, loss: 0.5782 Iter 4123, loss: 0.6178 Iter 4124, loss: 0.5688 Iter 4125, loss: 0.6001 Iter 4126, loss: 0.5750 Iter 4127, loss: 0.6232 Iter 4128, loss: 0.5661 Iter 4129, loss: 0.5807 Iter 4130, loss: 0.5993 Iter 4131, loss: 0.5637 Iter 4132, loss: 0.6155 Iter 4133, loss: 0.6001 Iter 4134, loss: 0.5475 Iter 4135, loss: 0.6068 Iter 4136, loss: 0.6305 Iter 4137, loss: 0.6288 Iter 4138, loss: 0.5795 Iter 4139, loss: 0.5628 Iter 4140, loss: 0.5763 Iter 4141, loss: 0.5796

Iter 4142, loss: 0.6311 Iter 4143, loss: 0.6352 Iter 4144, loss: 0.5932 Iter 4145, loss: 0.5720 Iter 4146, loss: 0.5638 Iter 4147, loss: 0.5871 Iter 4148, loss: 0.6190 Iter 4149, loss: 0.5772 Iter 4150, loss: 0.5909 Iter 4151, loss: 0.6015 Iter 4152, loss: 0.5684 Iter 4153, loss: 0.5838 Iter 4154, loss: 0.5564 Iter 4155, loss: 0.5991 Iter 4156, loss: 0.5573 Iter 4157, loss: 0.5883 Iter 4158, loss: 0.6367 Iter 4159, loss: 0.6178 Iter 4160, loss: 0.5844 Iter 4161, loss: 0.6336 Iter 4162, loss: 0.6202 Iter 4163, loss: 0.5806 Iter 4164, loss: 0.5857 Iter 4165, loss: 0.5946 Iter 4166, loss: 0.5950 Iter 4167, loss: 0.6092 Iter 4168, loss: 0.5673 Iter 4169, loss: 0.5510 Iter 4170, loss: 0.5742 Iter 4171, loss: 0.5980 Iter 4172, loss: 0.6585 Iter 4173, loss: 0.6386 Iter 4174, loss: 0.6013 Iter 4175, loss: 0.5918 Iter 4176, loss: 0.5783 Iter 4177, loss: 0.6104 Iter 4178, loss: 0.5491 Iter 4179, loss: 0.5705 Iter 4180, loss: 0.5589 Iter 4181, loss: 0.5845 Iter 4182, loss: 0.6040

Iter 4183, loss: 0.5829 Iter 4184, loss: 0.6223 Iter 4185, loss: 0.5915 Iter 4186, loss: 0.5919 Iter 4187, loss: 0.6077 Iter 4188, loss: 0.5972 Iter 4189, loss: 0.5616 Iter 4190, loss: 0.5675 Iter 4191, loss: 0.5664 Iter 4192, loss: 0.6382 Iter 4193, loss: 0.5945 Iter 4194, loss: 0.5903 Iter 4195, loss: 0.5306 Iter 4196, loss: 0.5844 Iter 4197, loss: 0.6232 Iter 4198, loss: 0.5831 Iter 4199, loss: 0.6177 Iter 4200, loss: 0.5812 Iter 4201, loss: 0.5571 Iter 4202, loss: 0.5776 Iter 4203, loss: 0.6007 Iter 4204, loss: 0.6195 Iter 4205, loss: 0.5504 Iter 4206, loss: 0.6220 Iter 4207, loss: 0.6112 Iter 4208, loss: 0.6021 Iter 4209, loss: 0.6010 Iter 4210, loss: 0.6128 Iter 4211, loss: 0.6200 Iter 4212, loss: 0.6140 Iter 4213, loss: 0.6080 Iter 4214, loss: 0.6341 Iter 4215, loss: 0.5582 Iter 4216, loss: 0.5717 Iter 4217, loss: 0.6276 Iter 4218, loss: 0.6050 Iter 4219, loss: 0.6199 Iter 4220, loss: 0.5718 Iter 4221, loss: 0.5713 Iter 4222, loss: 0.5266 Iter 4223, loss: 0.5329 Iter 4224, loss: 0.5815

Iter 4225, loss: 0.6036 Iter 4226, loss: 0.5718 Iter 4227, loss: 0.6067 Iter 4228, loss: 0.6115 Iter 4229, loss: 0.6120 Iter 4230, loss: 0.6095 Iter 4231, loss: 0.5818 Iter 4232, loss: 0.6164 Iter 4233, loss: 0.5973 Iter 4234, loss: 0.5904 Iter 4235, loss: 0.5863 Iter 4236, loss: 0.6105 Iter 4237, loss: 0.5862 Iter 4238, loss: 0.5674 Iter 4239, loss: 0.6265 Iter 4240, loss: 0.6008 Iter 4241, loss: 0.6092 Iter 4242, loss: 0.5892 Iter 4243, loss: 0.5995 Iter 4244, loss: 0.5965 Iter 4245, loss: 0.5930 Iter 4246, loss: 0.6172 Iter 4247, loss: 0.6104 Iter 4248, loss: 0.5639 Iter 4249, loss: 0.6010 Iter 4250, loss: 0.5907 Iter 4251, loss: 0.5987 Iter 4252, loss: 0.5688 Iter 4253, loss: 0.5885 Iter 4254, loss: 0.5731 Iter 4255, loss: 0.5203 Iter 4256, loss: 0.5944 Iter 4257, loss: 0.6040 Iter 4258, loss: 0.5910 Iter 4259, loss: 0.5454 Iter 4260, loss: 0.5438 Iter 4261, loss: 0.5757 Iter 4262, loss: 0.5806 Iter 4263, loss: 0.5927 Iter 4264, loss: 0.5734 Iter 4265, loss: 0.6093 Iter 4266, loss: 0.5508

Iter 4267, loss: 0.6663 Iter 4268, loss: 0.5712 Iter 4269, loss: 0.6118 Iter 4270, loss: 0.6076 Iter 4271, loss: 0.5600 Iter 4272, loss: 0.5815 Iter 4273, loss: 0.6170 Iter 4274, loss: 0.5672 Iter 4275, loss: 0.5525 Iter 4276, loss: 0.5482 Iter 4277, loss: 0.5768 Iter 4278, loss: 0.5832 Iter 4279, loss: 0.5974 Iter 4280, loss: 0.6121 Iter 4281, loss: 0.5730 Iter 4282, loss: 0.5975 Iter 4283, loss: 0.5819 Iter 4284, loss: 0.5675 Iter 4285, loss: 0.5873 Iter 4286, loss: 0.5944 Iter 4287, loss: 0.5860 Iter 4288, loss: 0.5832 Iter 4289, loss: 0.6087 Iter 4290, loss: 0.5897 Iter 4291, loss: 0.5631 Iter 4292, loss: 0.6027 Iter 4293, loss: 0.5756 Iter 4294, loss: 0.6312 Iter 4295, loss: 0.6192 Iter 4296, loss: 0.5649 Iter 4297, loss: 0.6098 Iter 4298, loss: 0.6054 Iter 4299, loss: 0.6002 Iter 4300, loss: 0.5820 Iter 4301, loss: 0.5912 Iter 4302, loss: 0.5939 Iter 4303, loss: 0.5214 Iter 4304, loss: 0.6204 Iter 4305, loss: 0.5693 Iter 4306, loss: 0.5860 Iter 4307, loss: 0.6633 Iter 4308, loss: 0.5803

Iter 4309, loss: 0.6544 Iter 4310, loss: 0.5886 Iter 4311, loss: 0.5630 Iter 4312, loss: 0.5537 Iter 4313, loss: 0.5953 Iter 4314, loss: 0.6256 Iter 4315, loss: 0.6026 Iter 4316, loss: 0.5534 Iter 4317, loss: 0.6131 Iter 4318, loss: 0.6210 Iter 4319, loss: 0.5559 Iter 4320, loss: 0.5714 Iter 4321, loss: 0.5900 Iter 4322, loss: 0.5789 Iter 4323, loss: 0.6083 Iter 4324, loss: 0.5875 Iter 4325, loss: 0.5946 Iter 4326, loss: 0.6099 Iter 4327, loss: 0.5963 Iter 4328, loss: 0.5589 Iter 4329, loss: 0.5943 Iter 4330, loss: 0.6063 Iter 4331, loss: 0.5808 Iter 4332, loss: 0.5587 Iter 4333, loss: 0.5776 Iter 4334, loss: 0.6258 Iter 4335, loss: 0.5892 Iter 4336, loss: 0.5964 Iter 4337, loss: 0.6124 Iter 4338, loss: 0.5694 Iter 4339, loss: 0.5506 Iter 4340, loss: 0.6032 Iter 4341, loss: 0.6001 Iter 4342, loss: 0.6140 Iter 4343, loss: 0.6010 Iter 4344, loss: 0.5895 Iter 4345, loss: 0.5990 Iter 4346, loss: 0.6142 Iter 4347, loss: 0.6144 Iter 4348, loss: 0.5795 Iter 4349, loss: 0.5830 Iter 4350, loss: 0.6104

Iter 4351, loss: 0.6415 Iter 4352, loss: 0.5684 Iter 4353, loss: 0.5808 Iter 4354, loss: 0.6090 Iter 4355, loss: 0.6150 Iter 4356, loss: 0.5758 Iter 4357, loss: 0.5722 Iter 4358, loss: 0.5739 Iter 4359, loss: 0.6104 Iter 4360, loss: 0.5998 Iter 4361, loss: 0.6077 Iter 4362, loss: 0.5769 Iter 4363, loss: 0.5977 Iter 4364, loss: 0.5793 Iter 4365, loss: 0.6116 Iter 4366, loss: 0.5724 Iter 4367, loss: 0.5658 Iter 4368, loss: 0.5955 Iter 4369, loss: 0.5607 Iter 4370, loss: 0.5961 Iter 4371, loss: 0.5923 Iter 4372, loss: 0.5978 Iter 4373, loss: 0.5939 Iter 4374, loss: 0.6431 Iter 4375, loss: 0.5646 Iter 4376, loss: 0.5929 Iter 4377, loss: 0.5863 Iter 4378, loss: 0.6080 Iter 4379, loss: 0.5722 Iter 4380, loss: 0.5977 Iter 4381, loss: 0.5585 Iter 4382, loss: 0.6113 Iter 4383, loss: 0.5706 Iter 4384, loss: 0.6458 Iter 4385, loss: 0.5957 Iter 4386, loss: 0.6251 Iter 4387, loss: 0.5743 Iter 4388, loss: 0.5764 Iter 4389, loss: 0.6146 Iter 4390, loss: 0.5786 Iter 4391, loss: 0.5998

Iter 4392, loss: 0.5943 Iter 4393, loss: 0.5531 Iter 4394, loss: 0.5600 Iter 4395, loss: 0.5631 Iter 4396, loss: 0.6539 Iter 4397, loss: 0.5684 Iter 4398, loss: 0.6346 Iter 4399, loss: 0.5637 Iter 4400, loss: 0.5681 Iter 4401, loss: 0.6329 Iter 4402, loss: 0.5665 Iter 4403, loss: 0.5944 Iter 4404, loss: 0.5967 Iter 4405, loss: 0.5843 Iter 4406, loss: 0.5593 Iter 4407, loss: 0.5293 Iter 4408, loss: 0.5885 Iter 4409, loss: 0.6142 Iter 4410, loss: 0.6416 Iter 4411, loss: 0.5822 Iter 4412, loss: 0.5556 Iter 4413, loss: 0.5718 Iter 4414, loss: 0.5694 Iter 4415, loss: 0.6205 Iter 4416, loss: 0.6095 Iter 4417, loss: 0.6205 Iter 4418, loss: 0.5911 Iter 4419, loss: 0.5693 Iter 4420, loss: 0.5581 Iter 4421, loss: 0.5628 Iter 4422, loss: 0.5686 Iter 4423, loss: 0.6007 Iter 4424, loss: 0.4896 Iter 4425, loss: 0.5934 Iter 4426, loss: 0.6464 Iter 4427, loss: 0.5746 Iter 4428, loss: 0.5866 Iter 4429, loss: 0.6017 Iter 4430, loss: 0.5763 Iter 4431, loss: 0.5547 Iter 4432, loss: 0.6280 Iter 4433, loss: 0.6438

Iter 4434, loss: 0.6229 Iter 4435, loss: 0.5746 Iter 4436, loss: 0.6295 Iter 4437, loss: 0.6284 Iter 4438, loss: 0.5616 Iter 4439, loss: 0.5597 Iter 4440, loss: 0.5945 Iter 4441, loss: 0.5769 Iter 4442, loss: 0.6299 Iter 4443, loss: 0.6092 Iter 4444, loss: 0.5949 Iter 4445, loss: 0.6113 Iter 4446, loss: 0.5832 Iter 4447, loss: 0.6033 Iter 4448, loss: 0.6119 Iter 4449, loss: 0.5992 Iter 4450, loss: 0.6125 Iter 4451, loss: 0.6071 Iter 4452, loss: 0.5742 Iter 4453, loss: 0.5991 Iter 4454, loss: 0.5832 Iter 4455, loss: 0.5772 Iter 4456, loss: 0.5684 Iter 4457, loss: 0.6398 Iter 4458, loss: 0.6104 Iter 4459, loss: 0.6105 Iter 4460, loss: 0.6029 Iter 4461, loss: 0.5924 Iter 4462, loss: 0.5925 Iter 4463, loss: 0.5885 Iter 4464, loss: 0.5817 Iter 4465, loss: 0.6003 Iter 4466, loss: 0.5859 Iter 4467, loss: 0.5796 Iter 4468, loss: 0.5829 Iter 4469, loss: 0.5271 Iter 4470, loss: 0.6108 Iter 4471, loss: 0.5953 Iter 4472, loss: 0.5750 Iter 4473, loss: 0.5917 Iter 4474, loss: 0.5701 Iter 4475, loss: 0.5790

Iter 4476, loss: 0.5409 Iter 4477, loss: 0.5750 Iter 4478, loss: 0.6002 Iter 4479, loss: 0.5501 Iter 4480, loss: 0.5658 Iter 4481, loss: 0.6353 Iter 4482, loss: 0.5590 Iter 4483, loss: 0.6051 Iter 4484, loss: 0.6099 Iter 4485, loss: 0.5818 Iter 4486, loss: 0.5943 Iter 4487, loss: 0.6341 Iter 4488, loss: 0.5413 Iter 4489, loss: 0.5982 Iter 4490, loss: 0.5698 Iter 4491, loss: 0.6422 Iter 4492, loss: 0.5483 Iter 4493, loss: 0.5908 Iter 4494, loss: 0.5747 Iter 4495, loss: 0.6323 Iter 4496, loss: 0.5757 Iter 4497, loss: 0.5542 Iter 4498, loss: 0.5862 Iter 4499, loss: 0.5788 Iter 4500, loss: 0.5916 Iter 4501, loss: 0.6328 Iter 4502, loss: 0.6195 Iter 4503, loss: 0.5669 Iter 4504, loss: 0.5733 Iter 4505, loss: 0.5946 Iter 4506, loss: 0.5346 Iter 4507, loss: 0.5475 Iter 4508, loss: 0.5631 Iter 4509, loss: 0.5801 Iter 4510, loss: 0.5815 Iter 4511, loss: 0.6077 Iter 4512, loss: 0.5558 Iter 4513, loss: 0.5596 Iter 4514, loss: 0.5850 Iter 4515, loss: 0.6144 Iter 4516, loss: 0.6161 Iter 4517, loss: 0.5755

Iter 4518, loss: 0.5857 Iter 4519, loss: 0.5873 Iter 4520, loss: 0.5912 Iter 4521, loss: 0.6179 Iter 4522, loss: 0.6050 Iter 4523, loss: 0.6162 Iter 4524, loss: 0.5898 Iter 4525, loss: 0.5850 Iter 4526, loss: 0.5463 Iter 4527, loss: 0.5513 Iter 4528, loss: 0.5839 Iter 4529, loss: 0.6051 Iter 4530, loss: 0.5947 Iter 4531, loss: 0.5844 Iter 4532, loss: 0.5609 Iter 4533, loss: 0.5567 Iter 4534, loss: 0.6067 Iter 4535, loss: 0.6064 Iter 4536, loss: 0.5775 Iter 4537, loss: 0.6117 Iter 4538, loss: 0.6239 Iter 4539, loss: 0.5755 Iter 4540, loss: 0.6235 Iter 4541, loss: 0.5673 Iter 4542, loss: 0.5804 Iter 4543, loss: 0.5985 Iter 4544, loss: 0.5583 Iter 4545, loss: 0.6119 Iter 4546, loss: 0.5824 Iter 4547, loss: 0.6342 Iter 4548, loss: 0.6478 Iter 4549, loss: 0.6051 Iter 4550, loss: 0.5946 Iter 4551, loss: 0.6184 Iter 4552, loss: 0.5881 Iter 4553, loss: 0.5556 Iter 4554, loss: 0.5824 Iter 4555, loss: 0.5742 Iter 4556, loss: 0.6442 Iter 4557, loss: 0.5654 Iter 4558, loss: 0.6005 Iter 4559, loss: 0.6169

Iter 4560, loss: 0.6160 Iter 4561, loss: 0.5911 Iter 4562, loss: 0.6274 Iter 4563, loss: 0.5954 Iter 4564, loss: 0.5589 Iter 4565, loss: 0.5671 Iter 4566, loss: 0.5532 Iter 4567, loss: 0.6030 Iter 4568, loss: 0.6088 Iter 4569, loss: 0.5622 Iter 4570, loss: 0.6075 Iter 4571, loss: 0.5921 Iter 4572, loss: 0.5926 Iter 4573, loss: 0.5504 Iter 4574, loss: 0.6406 Iter 4575, loss: 0.5833 Iter 4576, loss: 0.6259 Iter 4577, loss: 0.5968 Iter 4578, loss: 0.6332 Iter 4579, loss: 0.5887 Iter 4580, loss: 0.6335 Iter 4581, loss: 0.5727 Iter 4582, loss: 0.5973 Iter 4583, loss: 0.5425 Iter 4584, loss: 0.5987 Iter 4585, loss: 0.5923 Iter 4586, loss: 0.6279 Iter 4587, loss: 0.5829 Iter 4588, loss: 0.5970 Iter 4589, loss: 0.5411 Iter 4590, loss: 0.6191 Iter 4591, loss: 0.5764 Iter 4592, loss: 0.5582 Iter 4593, loss: 0.5998 Iter 4594, loss: 0.5694 Iter 4595, loss: 0.6111 Iter 4596, loss: 0.6059 Iter 4597, loss: 0.5808 Iter 4598, loss: 0.5703 Iter 4599, loss: 0.6002 Iter 4600, loss: 0.5646

Iter 4601, loss: 0.5928 Iter 4602, loss: 0.5599 Iter 4603, loss: 0.6020 Iter 4604, loss: 0.5714 Iter 4605, loss: 0.6082 Iter 4606, loss: 0.5833 Iter 4607, loss: 0.5839 Iter 4608, loss: 0.6487 Iter 4609, loss: 0.5883 Iter 4610, loss: 0.6046 Iter 4611, loss: 0.5715 Iter 4612, loss: 0.5848 Iter 4613, loss: 0.6109 Iter 4614, loss: 0.5912 Iter 4615, loss: 0.6324 Iter 4616, loss: 0.5521 Iter 4617, loss: 0.5849 Iter 4618, loss: 0.5918 Iter 4619, loss: 0.5944 Iter 4620, loss: 0.5579 Iter 4621, loss: 0.5936 Iter 4622, loss: 0.6025 Iter 4623, loss: 0.5763 Iter 4624, loss: 0.5877 Iter 4625, loss: 0.5622 Iter 4626, loss: 0.6039 Iter 4627, loss: 0.5749 Iter 4628, loss: 0.6220 Iter 4629, loss: 0.5829 Iter 4630, loss: 0.5991 Iter 4631, loss: 0.5776 Iter 4632, loss: 0.5263 Iter 4633, loss: 0.5871 Iter 4634, loss: 0.5845 Iter 4635, loss: 0.5948 Iter 4636, loss: 0.6040 Iter 4637, loss: 0.6311 Iter 4638, loss: 0.6324 Iter 4639, loss: 0.5691 Iter 4640, loss: 0.5273 Iter 4641, loss: 0.5818 Iter 4642, loss: 0.5716

Iter 4643, loss: 0.5951 Iter 4644, loss: 0.6219 Iter 4645, loss: 0.5849 Iter 4646, loss: 0.5772 Iter 4647, loss: 0.6120 Iter 4648, loss: 0.5863 Iter 4649, loss: 0.5401 Iter 4650, loss: 0.6018 Iter 4651, loss: 0.6109 Iter 4652, loss: 0.5908 Iter 4653, loss: 0.5393 Iter 4654, loss: 0.5779 Iter 4655, loss: 0.5727 Iter 4656, loss: 0.5977 Iter 4657, loss: 0.5999 Iter 4658, loss: 0.5605 Iter 4659, loss: 0.6023 Iter 4660, loss: 0.6290 Iter 4661, loss: 0.6165 Iter 4662, loss: 0.5602 Iter 4663, loss: 0.5687 Iter 4664, loss: 0.5900 Iter 4665, loss: 0.6082 Iter 4666, loss: 0.6156 Iter 4667, loss: 0.5672 Iter 4668, loss: 0.6023 Iter 4669, loss: 0.6051 Iter 4670, loss: 0.6284 Iter 4671, loss: 0.6016 Iter 4672, loss: 0.5488 Iter 4673, loss: 0.5613 Iter 4674, loss: 0.5859 Iter 4675, loss: 0.5750 Iter 4676, loss: 0.5536 Iter 4677, loss: 0.5921 Iter 4678, loss: 0.5974 Iter 4679, loss: 0.6197 Iter 4680, loss: 0.5878 Iter 4681, loss: 0.6291 Iter 4682, loss: 0.5937 Iter 4683, loss: 0.6205 Iter 4684, loss: 0.6192

Iter 4685, loss: 0.5946 Iter 4686, loss: 0.6119 Iter 4687, loss: 0.5900 Iter 4688, loss: 0.5919 Iter 4689, loss: 0.6003 Iter 4690, loss: 0.6141 Iter 4691, loss: 0.6321 Iter 4692, loss: 0.6113 Iter 4693, loss: 0.5902 Iter 4694, loss: 0.5707 Iter 4695, loss: 0.5483 Iter 4696, loss: 0.5836 Iter 4697, loss: 0.5673 Iter 4698, loss: 0.5480 Iter 4699, loss: 0.6131 Iter 4700, loss: 0.5771 Iter 4701, loss: 0.6041 Iter 4702, loss: 0.5892 Iter 4703, loss: 0.5709 Iter 4704, loss: 0.5983 Iter 4705, loss: 0.6102 Iter 4706, loss: 0.6008 Iter 4707, loss: 0.5466 Iter 4708, loss: 0.5511 Iter 4709, loss: 0.5913 Iter 4710, loss: 0.5967 Iter 4711, loss: 0.5989 Iter 4712, loss: 0.5177 Iter 4713, loss: 0.5995 Iter 4714, loss: 0.5745 Iter 4715, loss: 0.5411 Iter 4716, loss: 0.5661 Iter 4717, loss: 0.6058 Iter 4718, loss: 0.5535 Iter 4719, loss: 0.6350 Iter 4720, loss: 0.5906 Iter 4721, loss: 0.5906 Iter 4722, loss: 0.5864 Iter 4723, loss: 0.6186 Iter 4724, loss: 0.5832 Iter 4725, loss: 0.5947

Iter 4726, loss: 0.5965 Iter 4727, loss: 0.5725 Iter 4728, loss: 0.5761 Iter 4729, loss: 0.5639 Iter 4730, loss: 0.6005 Iter 4731, loss: 0.5795 Iter 4732, loss: 0.5501 Iter 4733, loss: 0.5919 Iter 4734, loss: 0.5992 Iter 4735, loss: 0.5694 Iter 4736, loss: 0.6007 Iter 4737, loss: 0.6121 Iter 4738, loss: 0.6052 Iter 4739, loss: 0.5959 Iter 4740, loss: 0.6103 Iter 4741, loss: 0.6390 Iter 4742, loss: 0.5648 Iter 4743, loss: 0.5963 Iter 4744, loss: 0.5397 Iter 4745, loss: 0.5800 Iter 4746, loss: 0.5833 Iter 4747, loss: 0.5507 Iter 4748, loss: 0.6330 Iter 4749, loss: 0.5795 Iter 4750, loss: 0.5837 Iter 4751, loss: 0.5880 Iter 4752, loss: 0.5764 Iter 4753, loss: 0.5794 Iter 4754, loss: 0.5667 Iter 4755, loss: 0.6326 Iter 4756, loss: 0.5682 Iter 4757, loss: 0.5635 Iter 4758, loss: 0.5746 Iter 4759, loss: 0.6066 Iter 4760, loss: 0.6140 Iter 4761, loss: 0.5752 Iter 4762, loss: 0.6122 Iter 4763, loss: 0.5472 Iter 4764, loss: 0.5805 Iter 4765, loss: 0.5994 Iter 4766, loss: 0.5836

Iter 4767, loss: 0.5731 Iter 4768, loss: 0.5974 Iter 4769, loss: 0.5619 Iter 4770, loss: 0.5645 Iter 4771, loss: 0.6020 Iter 4772, loss: 0.5909 Iter 4773, loss: 0.5824 Iter 4774, loss: 0.5765 Iter 4775, loss: 0.5854 Iter 4776, loss: 0.6128 Iter 4777, loss: 0.5886 Iter 4778, loss: 0.6054 Iter 4779, loss: 0.5640 Iter 4780, loss: 0.5699 Iter 4781, loss: 0.5804 Iter 4782, loss: 0.5632 Iter 4783, loss: 0.5913 Iter 4784, loss: 0.5886 Iter 4785, loss: 0.5916 Iter 4786, loss: 0.5660 Iter 4787, loss: 0.5619 Iter 4788, loss: 0.5610 Iter 4789, loss: 0.6063 Iter 4790, loss: 0.6013 Iter 4791, loss: 0.5630 Iter 4792, loss: 0.5770 Iter 4793, loss: 0.5904 Iter 4794, loss: 0.6047 Iter 4795, loss: 0.5586 Iter 4796, loss: 0.5948 Iter 4797, loss: 0.5861 Iter 4798, loss: 0.5952 Iter 4799, loss: 0.6317 Iter 4800, loss: 0.5378 Iter 4801, loss: 0.6096 Iter 4802, loss: 0.5755 Iter 4803, loss: 0.5447 Iter 4804, loss: 0.5580 Iter 4805, loss: 0.6145 Iter 4806, loss: 0.5885 Iter 4807, loss: 0.5770 Iter 4808, loss: 0.5741

Iter 4809, loss: 0.6078 Iter 4810, loss: 0.5610 Iter 4811, loss: 0.5607 Iter 4812, loss: 0.5131 Iter 4813, loss: 0.5932 Iter 4814, loss: 0.5679 Iter 4815, loss: 0.6088 Iter 4816, loss: 0.5703 Iter 4817, loss: 0.6100 Iter 4818, loss: 0.5952 Iter 4819, loss: 0.5401 Iter 4820, loss: 0.5692 Iter 4821, loss: 0.5710 Iter 4822, loss: 0.5562 Iter 4823, loss: 0.5642 Iter 4824, loss: 0.5992 Iter 4825, loss: 0.5668 Iter 4826, loss: 0.5836 Iter 4827, loss: 0.5627 Iter 4828, loss: 0.6078 Iter 4829, loss: 0.5878 Iter 4830, loss: 0.5944 Iter 4831, loss: 0.5763 Iter 4832, loss: 0.5716 Iter 4833, loss: 0.6130 Iter 4834, loss: 0.5822 Iter 4835, loss: 0.5927 Iter 4836, loss: 0.6156 Iter 4837, loss: 0.6156 Iter 4838, loss: 0.6001 Iter 4839, loss: 0.6048 Iter 4840, loss: 0.5705 Iter 4841, loss: 0.5552 Iter 4842, loss: 0.5790 Iter 4843, loss: 0.6200 Iter 4844, loss: 0.6045 Iter 4845, loss: 0.5321 Iter 4846, loss: 0.5641

Iter 4847, loss: 0.5708 Iter 4848, loss: 0.6081 Iter 4849, loss: 0.6001 Iter 4850, loss: 0.5720 Iter 4851, loss: 0.5476 Iter 4852, loss: 0.5939 Iter 4853, loss: 0.5543 Iter 4854, loss: 0.5575 Iter 4855, loss: 0.5959 Iter 4856, loss: 0.6117 Iter 4857, loss: 0.5978 Iter 4858, loss: 0.5841 Iter 4859, loss: 0.5572 Iter 4860, loss: 0.5433 Iter 4861, loss: 0.5541 Iter 4862, loss: 0.5615 Iter 4863, loss: 0.5590 Iter 4864, loss: 0.6147 Iter 4865, loss: 0.5876 Iter 4866, loss: 0.6106 Iter 4867, loss: 0.6101 Iter 4868, loss: 0.5742 Iter 4869, loss: 0.5909 Iter 4870, loss: 0.5998 Iter 4871, loss: 0.5866 Iter 4872, loss: 0.5734 Iter 4873, loss: 0.5969 Iter 4874, loss: 0.5820 Iter 4875, loss: 0.6107 Iter 4876, loss: 0.5989 Iter 4877, loss: 0.5894 Iter 4878, loss: 0.6086 Iter 4879, loss: 0.6480 Iter 4880, loss: 0.5848 Iter 4881, loss: 0.5824 Iter 4882, loss: 0.5870 Iter 4883, loss: 0.5378 Iter 4884, loss: 0.5992 Iter 4885, loss: 0.5970 Iter 4886, loss: 0.5699 Iter 4887, loss: 0.5665 Iter 4888, loss: 0.5472

Iter 4889, loss: 0.5683 Iter 4890, loss: 0.5816 Iter 4891, loss: 0.5793 Iter 4892, loss: 0.5961 Iter 4893, loss: 0.5961 Iter 4894, loss: 0.5835 Iter 4895, loss: 0.5791 Iter 4896, loss: 0.6147 Iter 4897, loss: 0.5748 Iter 4898, loss: 0.5846 Iter 4899, loss: 0.5963 Iter 4900, loss: 0.6276 Iter 4901, loss: 0.5263 Iter 4902, loss: 0.5656 Iter 4903, loss: 0.5923 Iter 4904, loss: 0.5444 Iter 4905, loss: 0.5814 Iter 4906, loss: 0.5755 Iter 4907, loss: 0.5614 Iter 4908, loss: 0.5680 Iter 4909, loss: 0.5978 Iter 4910, loss: 0.6176 Iter 4911, loss: 0.5870 Iter 4912, loss: 0.5836 Iter 4913, loss: 0.6253 Iter 4914, loss: 0.6215 Iter 4915, loss: 0.5871 Iter 4916, loss: 0.5838 Iter 4917, loss: 0.6055 Iter 4918, loss: 0.5969 Iter 4919, loss: 0.5581 Iter 4920, loss: 0.5392 Iter 4921, loss: 0.5937 Iter 4922, loss: 0.6031 Iter 4923, loss: 0.5941 Iter 4924, loss: 0.6037 Iter 4925, loss: 0.5847 Iter 4926, loss: 0.5691 Iter 4927, loss: 0.5571 Iter 4928, loss: 0.5372 Iter 4929, loss: 0.6239

Iter 4930, loss: 0.6021 Iter 4931, loss: 0.5173 Iter 4932, loss: 0.5619 Iter 4933, loss: 0.5924 Iter 4934, loss: 0.6209 Iter 4935, loss: 0.5880 Iter 4936, loss: 0.5779 Iter 4937, loss: 0.5979 Iter 4938, loss: 0.5993 Iter 4939, loss: 0.5697 Iter 4940, loss: 0.6025 Iter 4941, loss: 0.5574 Iter 4942, loss: 0.5703 Iter 4943, loss: 0.6344 Iter 4944, loss: 0.5728 Iter 4945, loss: 0.6138 Iter 4946, loss: 0.5607 Iter 4947, loss: 0.5634 Iter 4948, loss: 0.6035 Iter 4949, loss: 0.6338 Iter 4950, loss: 0.5691 Iter 4951, loss: 0.5929 Iter 4952, loss: 0.5786 Iter 4953, loss: 0.5725 Iter 4954, loss: 0.5939 Iter 4955, loss: 0.5896 Iter 4956, loss: 0.5543 Iter 4957, loss: 0.5743 Iter 4958, loss: 0.5573 Iter 4959, loss: 0.5794 Iter 4960, loss: 0.5842 Iter 4961, loss: 0.5805 Iter 4962, loss: 0.6214 Iter 4963, loss: 0.6045 Iter 4964, loss: 0.5672 Iter 4965, loss: 0.5867 Iter 4966, loss: 0.6174 Iter 4967, loss: 0.5611 Iter 4968, loss: 0.6002 Iter 4969, loss: 0.5905 Iter 4970, loss: 0.5971 Iter 4971, loss: 0.5899

Iter 4972, loss: 0.5982 Iter 4973, loss: 0.5880 Iter 4974, loss: 0.5841 Iter 4975, loss: 0.5889 Iter 4976, loss: 0.5666 Iter 4977, loss: 0.5671 Iter 4978, loss: 0.5602 Iter 4979, loss: 0.6183 Iter 4980, loss: 0.5842 Iter 4981, loss: 0.5668 Iter 4982, loss: 0.5810 Iter 4983, loss: 0.6156 Iter 4984, loss: 0.5545 Iter 4985, loss: 0.5546 Iter 4986, loss: 0.5714 Iter 4987, loss: 0.6425 Iter 4988, loss: 0.6278 Iter 4989, loss: 0.5800 Iter 4990, loss: 0.5830 Iter 4991, loss: 0.5787 Iter 4992, loss: 0.6052 Iter 4993, loss: 0.6010 Iter 4994, loss: 0.6152 Iter 4995, loss: 0.5496 Iter 4996, loss: 0.6302 Iter 4997, loss: 0.5812 Iter 4998, loss: 0.6171 Iter 4999, loss: 0.5677 Iter 5000, loss: 0.5961 Iter 5001, loss: 0.5955 Iter 5002, loss: 0.5743 Iter 5003, loss: 0.6051 Iter 5004, loss: 0.5977 Iter 5005, loss: 0.5675 Iter 5006, loss: 0.5856 Iter 5007, loss: 0.5890 Iter 5008, loss: 0.6120 Iter 5009, loss: 0.5768 Iter 5010, loss: 0.5736 Iter 5011, loss: 0.5921 Iter 5012, loss: 0.6072 Iter 5013, loss: 0.6024

Iter 5014, loss: 0.6072 Iter 5015, loss: 0.6007 Iter 5016, loss: 0.6058 Iter 5017, loss: 0.5795 Iter 5018, loss: 0.5742 Iter 5019, loss: 0.5925 Iter 5020, loss: 0.5821 Iter 5021, loss: 0.5834 Iter 5022, loss: 0.5463 Iter 5023, loss: 0.5764 Iter 5024, loss: 0.6081 Iter 5025, loss: 0.5589 Iter 5026, loss: 0.5757 Iter 5027, loss: 0.5761 Iter 5028, loss: 0.6142 Iter 5029, loss: 0.5559 Iter 5030, loss: 0.5946 Iter 5031, loss: 0.5830 Iter 5032, loss: 0.5725 Iter 5033, loss: 0.5672 Iter 5034, loss: 0.6084 Iter 5035, loss: 0.6239 Iter 5036, loss: 0.6130 Iter 5037, loss: 0.5620 Iter 5038, loss: 0.5935 Iter 5039, loss: 0.5423 Iter 5040, loss: 0.5627 Iter 5041, loss: 0.5431 Iter 5042, loss: 0.5808 Iter 5043, loss: 0.5655 Iter 5044, loss: 0.6666 Iter 5045, loss: 0.5613 Iter 5046, loss: 0.5974 Iter 5047, loss: 0.6041 Iter 5048, loss: 0.5914 Iter 5049, loss: 0.6541 Iter 5050, loss: 0.5854 Iter 5051, loss: 0.5733 Iter 5052, loss: 0.5528 Iter 5053, loss: 0.5870 Iter 5054, loss: 0.5866

Iter 5055, loss: 0.6525 Iter 5056, loss: 0.5758 Iter 5057, loss: 0.5766 Iter 5058, loss: 0.6419 Iter 5059, loss: 0.5869 Iter 5060, loss: 0.5721 Iter 5061, loss: 0.6047 Iter 5062, loss: 0.6028 Iter 5063, loss: 0.5610 Iter 5064, loss: 0.5911 Iter 5065, loss: 0.6058 Iter 5066, loss: 0.5630 Iter 5067, loss: 0.5507 Iter 5068, loss: 0.5270 Iter 5069, loss: 0.5684 Iter 5070, loss: 0.6080 Iter 5071, loss: 0.5750 Iter 5072, loss: 0.5746 Iter 5073, loss: 0.5656 Iter 5074, loss: 0.5926 Iter 5075, loss: 0.5705 Iter 5076, loss: 0.6160 Iter 5077, loss: 0.5576 Iter 5078, loss: 0.5163 Iter 5079, loss: 0.5890 Iter 5080, loss: 0.5855 Iter 5081, loss: 0.5697 Iter 5082, loss: 0.5654 Iter 5083, loss: 0.5983 Iter 5084, loss: 0.5833 Iter 5085, loss: 0.5789 Iter 5086, loss: 0.5833 Iter 5087, loss: 0.5573 Iter 5088, loss: 0.5767 Iter 5089, loss: 0.5624 Iter 5090, loss: 0.6256 Iter 5091, loss: 0.6136 Iter 5092, loss: 0.5707 Iter 5093, loss: 0.5911 Iter 5094, loss: 0.5362 Iter 5095, loss: 0.5921

Iter 5096, loss: 0.5223 Iter 5097, loss: 0.5513 Iter 5098, loss: 0.5780 Iter 5099, loss: 0.5731 Iter 5100, loss: 0.5874 Iter 5101, loss: 0.6142 Iter 5102, loss: 0.6043 Iter 5103, loss: 0.5519 Iter 5104, loss: 0.6382 Iter 5105, loss: 0.5883 Iter 5106, loss: 0.5571 Iter 5107, loss: 0.6066 Iter 5108, loss: 0.5582 Iter 5109, loss: 0.5357 Iter 5110, loss: 0.5869 Iter 5111, loss: 0.5853 Iter 5112, loss: 0.5396 Iter 5113, loss: 0.6082 Iter 5114, loss: 0.5709 Iter 5115, loss: 0.5960 Iter 5116, loss: 0.5916 Iter 5117, loss: 0.5624 Iter 5118, loss: 0.5873 Iter 5119, loss: 0.5706 Iter 5120, loss: 0.5693 Iter 5121, loss: 0.5853 Iter 5122, loss: 0.5797 Iter 5123, loss: 0.5490 Iter 5124, loss: 0.5828 Iter 5125, loss: 0.5701 Iter 5126, loss: 0.5641 Iter 5127, loss: 0.6224 Iter 5128, loss: 0.5687 Iter 5129, loss: 0.5645 Iter 5130, loss: 0.5893 Iter 5131, loss: 0.5874 Iter 5132, loss: 0.5866 Iter 5133, loss: 0.5937 Iter 5134, loss: 0.5576 Iter 5135, loss: 0.5779 Iter 5136, loss: 0.5953 Iter 5137, loss: 0.6188

Iter 5138, loss: 0.5793 Iter 5139, loss: 0.5614 Iter 5140, loss: 0.5498 Iter 5141, loss: 0.5652 Iter 5142, loss: 0.6004 Iter 5143, loss: 0.5851 Iter 5144, loss: 0.5629 Iter 5145, loss: 0.5989 Iter 5146, loss: 0.5824 Iter 5147, loss: 0.5738 Iter 5148, loss: 0.6035 Iter 5149, loss: 0.5602 Iter 5150, loss: 0.5759 Iter 5151, loss: 0.5765 Iter 5152, loss: 0.5772 Iter 5153, loss: 0.5875 Iter 5154, loss: 0.5910 Iter 5155, loss: 0.5195 Iter 5156, loss: 0.5779 Iter 5157, loss: 0.5324 Iter 5158, loss: 0.5780 Iter 5159, loss: 0.5851 Iter 5160, loss: 0.5894 Iter 5161, loss: 0.5137 Iter 5162, loss: 0.6206 Iter 5163, loss: 0.6005 Iter 5164, loss: 0.5848 Iter 5165, loss: 0.5984 Iter 5166, loss: 0.5717 Iter 5167, loss: 0.6155 Iter 5168, loss: 0.5881 Iter 5169, loss: 0.5680 Iter 5170, loss: 0.5942 Iter 5171, loss: 0.5240 Iter 5172, loss: 0.5575 Iter 5173, loss: 0.5617 Iter 5174, loss: 0.5796 Iter 5175, loss: 0.5780 Iter 5176, loss: 0.5946 Iter 5177, loss: 0.5682 Iter 5178, loss: 0.5631 Iter 5179, loss: 0.6103

Iter 5180, loss: 0.5369 Iter 5181, loss: 0.5957 Iter 5182, loss: 0.5401 Iter 5183, loss: 0.6103 Iter 5184, loss: 0.5886 Iter 5185, loss: 0.5306 Iter 5186, loss: 0.5649 Iter 5187, loss: 0.6076 Iter 5188, loss: 0.6112 Iter 5189, loss: 0.5990 Iter 5190, loss: 0.5960 Iter 5191, loss: 0.5662 Iter 5192, loss: 0.5772 Iter 5193, loss: 0.5584 Iter 5194, loss: 0.5583 Iter 5195, loss: 0.5922 Iter 5196, loss: 0.5661 Iter 5197, loss: 0.6054 Iter 5198, loss: 0.6218 Iter 5199, loss: 0.6052 Iter 5200, loss: 0.5824 Iter 5201, loss: 0.5909 Iter 5202, loss: 0.5756 Iter 5203, loss: 0.5459 Iter 5204, loss: 0.5562 Iter 5205, loss: 0.5558 Iter 5206, loss: 0.6231 Iter 5207, loss: 0.6351 Iter 5208, loss: 0.5488 Iter 5209, loss: 0.5742 Iter 5210, loss: 0.5716 Iter 5211, loss: 0.5542 Iter 5212, loss: 0.6097 Iter 5213, loss: 0.6029 Iter 5214, loss: 0.5441 Iter 5215, loss: 0.6210 Iter 5216, loss: 0.5287 Iter 5217, loss: 0.6070 Iter 5218, loss: 0.5554 Iter 5219, loss: 0.5942 Iter 5220, loss: 0.6287 Iter 5221, loss: 0.5936

Iter 5222, loss: 0.5837 Iter 5223, loss: 0.5662 Iter 5224, loss: 0.5894 Iter 5225, loss: 0.5727 Iter 5226, loss: 0.5934 Iter 5227, loss: 0.6135 Iter 5228, loss: 0.5896 Iter 5229, loss: 0.5446 Iter 5230, loss: 0.5972 Iter 5231, loss: 0.6107 Iter 5232, loss: 0.5775 Iter 5233, loss: 0.5591 Iter 5234, loss: 0.5533 Iter 5235, loss: 0.6212 Iter 5236, loss: 0.5999 Iter 5237, loss: 0.5398 Iter 5238, loss: 0.5953 Iter 5239, loss: 0.5908 Iter 5240, loss: 0.5813 Iter 5241, loss: 0.5976 Iter 5242, loss: 0.6279 Iter 5243, loss: 0.5712 Iter 5244, loss: 0.6170 Iter 5245, loss: 0.5827 Iter 5246, loss: 0.5660 Iter 5247, loss: 0.5727 Iter 5248, loss: 0.5760 Iter 5249, loss: 0.5939 Iter 5250, loss: 0.5738 Iter 5251, loss: 0.5645 Iter 5252, loss: 0.6102 Iter 5253, loss: 0.5865 Iter 5254, loss: 0.5991 Iter 5255, loss: 0.5992 Iter 5256, loss: 0.5748 Iter 5257, loss: 0.5854 Iter 5258, loss: 0.5629 Iter 5259, loss: 0.5752 Iter 5260, loss: 0.5404 Iter 5261, loss: 0.5553 Iter 5262, loss: 0.5515

Iter 5263, loss: 0.5969 Iter 5264, loss: 0.5909 Iter 5265, loss: 0.6003 Iter 5266, loss: 0.5797 Iter 5267, loss: 0.5560 Iter 5268, loss: 0.5729 Iter 5269, loss: 0.5960 Iter 5270, loss: 0.5754 Iter 5271, loss: 0.5746 Iter 5272, loss: 0.5863 Iter 5273, loss: 0.5705 Iter 5274, loss: 0.5376 Iter 5275, loss: 0.5568 Iter 5276, loss: 0.5576 Iter 5277, loss: 0.5496 Iter 5278, loss: 0.5460 Iter 5279, loss: 0.5794 Iter 5280, loss: 0.5901 Iter 5281, loss: 0.6015 Iter 5282, loss: 0.5500 Iter 5283, loss: 0.5867 Iter 5284, loss: 0.5974 Iter 5285, loss: 0.6120 Iter 5286, loss: 0.5408 Iter 5287, loss: 0.5865 Iter 5288, loss: 0.5503 Iter 5289, loss: 0.5204 Iter 5290, loss: 0.5677 Iter 5291, loss: 0.5806 Iter 5292, loss: 0.5762 Iter 5293, loss: 0.5423 Iter 5294, loss: 0.5925 Iter 5295, loss: 0.6109 Iter 5296, loss: 0.5801 Iter 5297, loss: 0.5284 Iter 5298, loss: 0.5791 Iter 5299, loss: 0.6094 Iter 5300, loss: 0.6377 Iter 5301, loss: 0.5607 Iter 5302, loss: 0.5784 Iter 5303, loss: 0.5578 Iter 5304, loss: 0.5728

Iter 5305, loss: 0.6019 Iter 5306, loss: 0.5935 Iter 5307, loss: 0.5444 Iter 5308, loss: 0.5554 Iter 5309, loss: 0.5899 Iter 5310, loss: 0.5451 Iter 5311, loss: 0.5721 Iter 5312, loss: 0.5778 Iter 5313, loss: 0.5962 Iter 5314, loss: 0.5923 Iter 5315, loss: 0.5581 Iter 5316, loss: 0.6019 Iter 5317, loss: 0.5953 Iter 5318, loss: 0.5620 Iter 5319, loss: 0.5709 Iter 5320, loss: 0.6201 Iter 5321, loss: 0.6122 Iter 5322, loss: 0.5659 Iter 5323, loss: 0.5521 Iter 5324, loss: 0.6051 Iter 5325, loss: 0.6032 Iter 5326, loss: 0.6059 Iter 5327, loss: 0.5827 Iter 5328, loss: 0.5528 Iter 5329, loss: 0.5949 Iter 5330, loss: 0.5726 Iter 5331, loss: 0.5551 Iter 5332, loss: 0.5514 Iter 5333, loss: 0.5759 Iter 5334, loss: 0.5581 Iter 5335, loss: 0.5750 Iter 5336, loss: 0.5526 Iter 5337, loss: 0.5927 Iter 5338, loss: 0.5720 Iter 5339, loss: 0.5383 Iter 5340, loss: 0.5795 Iter 5341, loss: 0.5513 Iter 5342, loss: 0.5518 Iter 5343, loss: 0.5631 Iter 5344, loss: 0.5901 Iter 5345, loss: 0.5805 Iter 5346, loss: 0.5640

Iter 5347, loss: 0.6262 Iter 5348, loss: 0.5713 Iter 5349, loss: 0.5835 Iter 5350, loss: 0.5936 Iter 5351, loss: 0.5939 Iter 5352, loss: 0.5828 Iter 5353, loss: 0.5711 Iter 5354, loss: 0.5763 Iter 5355, loss: 0.5624 Iter 5356, loss: 0.5686 Iter 5357, loss: 0.5899 Iter 5358, loss: 0.6061 Iter 5359, loss: 0.5846 Iter 5360, loss: 0.5623 Iter 5361, loss: 0.6059 Iter 5362, loss: 0.5928 Iter 5363, loss: 0.5622 Iter 5364, loss: 0.5780 Iter 5365, loss: 0.5792 Iter 5366, loss: 0.5818 Iter 5367, loss: 0.5826 Iter 5368, loss: 0.5767 Iter 5369, loss: 0.5396 Iter 5370, loss: 0.5574 Iter 5371, loss: 0.5388 Iter 5372, loss: 0.5640 Iter 5373, loss: 0.5466 Iter 5374, loss: 0.5551 Iter 5375, loss: 0.5583 Iter 5376, loss: 0.5951 Iter 5377, loss: 0.5519 Iter 5378, loss: 0.5390 Iter 5379, loss: 0.5741 Iter 5380, loss: 0.5866 Iter 5381, loss: 0.5498 Iter 5382, loss: 0.5624 Iter 5383, loss: 0.5761 Iter 5384, loss: 0.6107 Iter 5385, loss: 0.6013 Iter 5386, loss: 0.5695 Iter 5387, loss: 0.5500

Iter 5388, loss: 0.5847 Iter 5389, loss: 0.5923 Iter 5390, loss: 0.5617 Iter 5391, loss: 0.5816 Iter 5392, loss: 0.6124 Iter 5393, loss: 0.6160 Iter 5394, loss: 0.5881 Iter 5395, loss: 0.5425 Iter 5396, loss: 0.6212 Iter 5397, loss: 0.5457 Iter 5398, loss: 0.5718 Iter 5399, loss: 0.6038 Iter 5400, loss: 0.6045 Iter 5401, loss: 0.5647 Iter 5402, loss: 0.6531 Iter 5403, loss: 0.5931 Iter 5404, loss: 0.6125 Iter 5405, loss: 0.5767 Iter 5406, loss: 0.6374 Iter 5407, loss: 0.5776 Iter 5408, loss: 0.5780 Iter 5409, loss: 0.5494 Iter 5410, loss: 0.5820 Iter 5411, loss: 0.5801 Iter 5412, loss: 0.5799 Iter 5413, loss: 0.6001 Iter 5414, loss: 0.5228 Iter 5415, loss: 0.5786 Iter 5416, loss: 0.5764 Iter 5417, loss: 0.5588 Iter 5418, loss: 0.5688 Iter 5419, loss: 0.5750 Iter 5420, loss: 0.5794 Iter 5421, loss: 0.5663 Iter 5422, loss: 0.5934 Iter 5423, loss: 0.5482 Iter 5424, loss: 0.5579 Iter 5425, loss: 0.5435 Iter 5426, loss: 0.5682 Iter 5427, loss: 0.5518 Iter 5428, loss: 0.5329

Iter 5429, loss: 0.5697 Iter 5430, loss: 0.5889 Iter 5431, loss: 0.5503 Iter 5432, loss: 0.5265 Iter 5433, loss: 0.6154 Iter 5434, loss: 0.6021 Iter 5435, loss: 0.5689 Iter 5436, loss: 0.5787 Iter 5437, loss: 0.5621 Iter 5438, loss: 0.5834 Iter 5439, loss: 0.5571 Iter 5440, loss: 0.5712 Iter 5441, loss: 0.5809 Iter 5442, loss: 0.5546 Iter 5443, loss: 0.5861 Iter 5444, loss: 0.6255 Iter 5445, loss: 0.5849 Iter 5446, loss: 0.5207 Iter 5447, loss: 0.5845 Iter 5448, loss: 0.5816 Iter 5449, loss: 0.5408 Iter 5450, loss: 0.6031 Iter 5451, loss: 0.6335 Iter 5452, loss: 0.6139 Iter 5453, loss: 0.6174 Iter 5454, loss: 0.6157 Iter 5455, loss: 0.5803 Iter 5456, loss: 0.5901 Iter 5457, loss: 0.5568 Iter 5458, loss: 0.6119 Iter 5459, loss: 0.6058 Iter 5460, loss: 0.5589 Iter 5461, loss: 0.5885 Iter 5462, loss: 0.5519 Iter 5463, loss: 0.5930 Iter 5464, loss: 0.5676 Iter 5465, loss: 0.5560 Iter 5466, loss: 0.5772 Iter 5467, loss: 0.5971 Iter 5468, loss: 0.6171 Iter 5469, loss: 0.5948 Iter 5470, loss: 0.5581

Iter 5471, loss: 0.5837 Iter 5472, loss: 0.5768 Iter 5473, loss: 0.5879 Iter 5474, loss: 0.5917 Iter 5475, loss: 0.6024 Iter 5476, loss: 0.5698 Iter 5477, loss: 0.5820 Iter 5478, loss: 0.5484 Iter 5479, loss: 0.5586 Iter 5480, loss: 0.5758 Iter 5481, loss: 0.5754 Iter 5482, loss: 0.5866 Iter 5483, loss: 0.5967 Iter 5484, loss: 0.5905 Iter 5485, loss: 0.5837 Iter 5486, loss: 0.5693 Iter 5487, loss: 0.5972 Iter 5488, loss: 0.5729 Iter 5489, loss: 0.5915 Iter 5490, loss: 0.5667 Iter 5491, loss: 0.5365 Iter 5492, loss: 0.6024 Iter 5493, loss: 0.5658 Iter 5494, loss: 0.5828 Iter 5495, loss: 0.6202 Iter 5496, loss: 0.5824 Iter 5497, loss: 0.6017 Iter 5498, loss: 0.5701 Iter 5499, loss: 0.5677 Iter 5500, loss: 0.5475 Iter 5501, loss: 0.5028 Iter 5502, loss: 0.6054 Iter 5503, loss: 0.5512 Iter 5504, loss: 0.5434 Iter 5505, loss: 0.5806 Iter 5506, loss: 0.6006 Iter 5507, loss: 0.5651 Iter 5508, loss: 0.5754 Iter 5509, loss: 0.5664 Iter 5510, loss: 0.5393 Iter 5511, loss: 0.5267

Iter 5512, loss: 0.5718 Iter 5513, loss: 0.5670 Iter 5514, loss: 0.5826 Iter 5515, loss: 0.5683 Iter 5516, loss: 0.5680 Iter 5517, loss: 0.5621 Iter 5518, loss: 0.5703 Iter 5519, loss: 0.6250 Iter 5520, loss: 0.5685 Iter 5521, loss: 0.5517 Iter 5522, loss: 0.5642 Iter 5523, loss: 0.5805 Iter 5524, loss: 0.5805 Iter 5525, loss: 0.6290 Iter 5526, loss: 0.5811 Iter 5527, loss: 0.5742 Iter 5528, loss: 0.5832 Iter 5529, loss: 0.5943 Iter 5530, loss: 0.5974 Iter 5531, loss: 0.5837 Iter 5532, loss: 0.5686 Iter 5533, loss: 0.5624 Iter 5534, loss: 0.5680 Iter 5535, loss: 0.5612 Iter 5536, loss: 0.5492 Iter 5537, loss: 0.5484 Iter 5538, loss: 0.6002 Iter 5539, loss: 0.5668 Iter 5540, loss: 0.5647 Iter 5541, loss: 0.6222 Iter 5542, loss: 0.6178 Iter 5543, loss: 0.5759 Iter 5544, loss: 0.5661 Iter 5545, loss: 0.5896 Iter 5546, loss: 0.5964 Iter 5547, loss: 0.5753 Iter 5548, loss: 0.5778 Iter 5549, loss: 0.6127 Iter 5550, loss: 0.5822 Iter 5551, loss: 0.5743 Iter 5552, loss: 0.5501 Iter 5553, loss: 0.5719

Iter 5554, loss: 0.5767 Iter 5555, loss: 0.5644 Iter 5556, loss: 0.5695 Iter 5557, loss: 0.6165 Iter 5558, loss: 0.5472 Iter 5559, loss: 0.5142 Iter 5560, loss: 0.5862 Iter 5561, loss: 0.5998 Iter 5562, loss: 0.5957 Iter 5563, loss: 0.5831 Iter 5564, loss: 0.5696 Iter 5565, loss: 0.5479 Iter 5566, loss: 0.5373 Iter 5567, loss: 0.5858 Iter 5568, loss: 0.5790 Iter 5569, loss: 0.5834 Iter 5570, loss: 0.5439 Iter 5571, loss: 0.5813 Iter 5572, loss: 0.5993 Iter 5573, loss: 0.5882 Iter 5574, loss: 0.5824 Iter 5575, loss: 0.5998 Iter 5576, loss: 0.5729 Iter 5577, loss: 0.5479 Iter 5578, loss: 0.6030 Iter 5579, loss: 0.5548 Iter 5580, loss: 0.5620 Iter 5581, loss: 0.5925 Iter 5582, loss: 0.6018 Iter 5583, loss: 0.5331 Iter 5584, loss: 0.5994 Iter 5585, loss: 0.5629 Iter 5586, loss: 0.5305 Iter 5587, loss: 0.5577 Iter 5588, loss: 0.5784 Iter 5589, loss: 0.5533 Iter 5590, loss: 0.5778 Iter 5591, loss: 0.6041 Iter 5592, loss: 0.6095 Iter 5593, loss: 0.6170 Iter 5594, loss: 0.5805

Iter 5595, loss: 0.5573 Iter 5596, loss: 0.5325 Iter 5597, loss: 0.5605 Iter 5598, loss: 0.6023 Iter 5599, loss: 0.5470 Iter 5600, loss: 0.5774 Iter 5601, loss: 0.5536 Iter 5602, loss: 0.6492 Iter 5603, loss: 0.6155 Iter 5604, loss: 0.5647 Iter 5605, loss: 0.5864 Iter 5606, loss: 0.5875 Iter 5607, loss: 0.5711 Iter 5608, loss: 0.5527 Iter 5609, loss: 0.6213 Iter 5610, loss: 0.5422 Iter 5611, loss: 0.5737 Iter 5612, loss: 0.5946 Iter 5613, loss: 0.5955 Iter 5614, loss: 0.5574 Iter 5615, loss: 0.6158 Iter 5616, loss: 0.6069 Iter 5617, loss: 0.5359 Iter 5618, loss: 0.5691 Iter 5619, loss: 0.5528 Iter 5620, loss: 0.5680 Iter 5621, loss: 0.5580 Iter 5622, loss: 0.5865 Iter 5623, loss: 0.5730 Iter 5624, loss: 0.5829 Iter 5625, loss: 0.5724 Iter 5626, loss: 0.5597 Iter 5627, loss: 0.6099 Iter 5628, loss: 0.5622 Iter 5629, loss: 0.5619 Iter 5630, loss: 0.5907 Iter 5631, loss: 0.5383 Iter 5632, loss: 0.5741 Iter 5633, loss: 0.5540 Iter 5634, loss: 0.5539 Iter 5635, loss: 0.5905 Iter 5636, loss: 0.5952

Iter 5637, loss: 0.5737 Iter 5638, loss: 0.5729 Iter 5639, loss: 0.6063 Iter 5640, loss: 0.6122 Iter 5641, loss: 0.5730 Iter 5642, loss: 0.5490 Iter 5643, loss: 0.6091 Iter 5644, loss: 0.5399 Iter 5645, loss: 0.5658 Iter 5646, loss: 0.5877 Iter 5647, loss: 0.5412 Iter 5648, loss: 0.5381 Iter 5649, loss: 0.5655 Iter 5650, loss: 0.5587 Iter 5651, loss: 0.5657 Iter 5652, loss: 0.5872 Iter 5653, loss: 0.5813 Iter 5654, loss: 0.5933 Iter 5655, loss: 0.5668 Iter 5656, loss: 0.6021 Iter 5657, loss: 0.6147 Iter 5658, loss: 0.6518 Iter 5659, loss: 0.6157 Iter 5660, loss: 0.5723 Iter 5661, loss: 0.6145 Iter 5662, loss: 0.6253 Iter 5663, loss: 0.5743 Iter 5664, loss: 0.5253 Iter 5665, loss: 0.5918 Iter 5666, loss: 0.5453 Iter 5667, loss: 0.5263 Iter 5668, loss: 0.5831 Iter 5669, loss: 0.5412 Iter 5670, loss: 0.5961 Iter 5671, loss: 0.5655 Iter 5672, loss: 0.5305 Iter 5673, loss: 0.5677 Iter 5674, loss: 0.5854 Iter 5675, loss: 0.5974 Iter 5676, loss: 0.6005 Iter 5677, loss: 0.5540 Iter 5678, loss: 0.5625

Iter 5679, loss: 0.5571 Iter 5680, loss: 0.5569 Iter 5681, loss: 0.6062 Iter 5682, loss: 0.6036 Iter 5683, loss: 0.6171 Iter 5684, loss: 0.5765 Iter 5685, loss: 0.5491 Iter 5686, loss: 0.5943 Iter 5687, loss: 0.5748 Iter 5688, loss: 0.5977 Iter 5689, loss: 0.5781 Iter 5690, loss: 0.5733 Iter 5691, loss: 0.5508 Iter 5692, loss: 0.5995 Iter 5693, loss: 0.5645 Iter 5694, loss: 0.5652 Iter 5695, loss: 0.6007 Iter 5696, loss: 0.6154 Iter 5697, loss: 0.5949 Iter 5698, loss: 0.5711 Iter 5699, loss: 0.5082 Iter 5700, loss: 0.5759 Iter 5701, loss: 0.5926 Iter 5702, loss: 0.5688 Iter 5703, loss: 0.5471 Iter 5704, loss: 0.5477 Iter 5705, loss: 0.5780 Iter 5706, loss: 0.5938 Iter 5707, loss: 0.5826 Iter 5708, loss: 0.5973 Iter 5709, loss: 0.5978 Iter 5710, loss: 0.5880 Iter 5711, loss: 0.6169 Iter 5712, loss: 0.5818 Iter 5713, loss: 0.6394 Iter 5714, loss: 0.5886 Iter 5715, loss: 0.6092 Iter 5716, loss: 0.5584 Iter 5717, loss: 0.5480 Iter 5718, loss: 0.5486 Iter 5719, loss: 0.6299

Iter 5720, loss: 0.5609 Iter 5721, loss: 0.5897 Iter 5722, loss: 0.6008 Iter 5723, loss: 0.5946 Iter 5724, loss: 0.5871 Iter 5725, loss: 0.5752 Iter 5726, loss: 0.5627 Iter 5727, loss: 0.5595 Iter 5728, loss: 0.5695 Iter 5729, loss: 0.5602 Iter 5730, loss: 0.6081 Iter 5731, loss: 0.5772 Iter 5732, loss: 0.5696 Iter 5733, loss: 0.5658 Iter 5734, loss: 0.5406 Iter 5735, loss: 0.5453 Iter 5736, loss: 0.5475 Iter 5737, loss: 0.5442 Iter 5738, loss: 0.5626 Iter 5739, loss: 0.5689 Iter 5740, loss: 0.6046 Iter 5741, loss: 0.5715 Iter 5742, loss: 0.5505 Iter 5743, loss: 0.5862 Iter 5744, loss: 0.5881 Iter 5745, loss: 0.5539 Iter 5746, loss: 0.5296 Iter 5747, loss: 0.5401 Iter 5748, loss: 0.5789 Iter 5749, loss: 0.5768 Iter 5750, loss: 0.5839 Iter 5751, loss: 0.5351 Iter 5752, loss: 0.5335 Iter 5753, loss: 0.6048 Iter 5754, loss: 0.5805 Iter 5755, loss: 0.5567 Iter 5756, loss: 0.5546 Iter 5757, loss: 0.5490 Iter 5758, loss: 0.5492 Iter 5759, loss: 0.5489 Iter 5760, loss: 0.5161

Iter 5761, loss: 0.5513 Iter 5762, loss: 0.6116 Iter 5763, loss: 0.5737 Iter 5764, loss: 0.5545 Iter 5765, loss: 0.5667 Iter 5766, loss: 0.6315 Iter 5767, loss: 0.6363 Iter 5768, loss: 0.5611 Iter 5769, loss: 0.5412 Iter 5770, loss: 0.5856 Iter 5771, loss: 0.5923 Iter 5772, loss: 0.6161 Iter 5773, loss: 0.5564 Iter 5774, loss: 0.5914 Iter 5775, loss: 0.5841 Iter 5776, loss: 0.5629 Iter 5777, loss: 0.5331 Iter 5778, loss: 0.5797 Iter 5779, loss: 0.5608 Iter 5780, loss: 0.5922 Iter 5781, loss: 0.5553 Iter 5782, loss: 0.5641 Iter 5783, loss: 0.6065 Iter 5784, loss: 0.6060 Iter 5785, loss: 0.5760 Iter 5786, loss: 0.5490 Iter 5787, loss: 0.5675 Iter 5788, loss: 0.6069 Iter 5789, loss: 0.5545 Iter 5790, loss: 0.5317 Iter 5791, loss: 0.5729 Iter 5792, loss: 0.5898 Iter 5793, loss: 0.5776 Iter 5794, loss: 0.5884 Iter 5795, loss: 0.5986 Iter 5796, loss: 0.6307 Iter 5797, loss: 0.6336 Iter 5798, loss: 0.6014 Iter 5799, loss: 0.5991 Iter 5800, loss: 0.5779 Iter 5801, loss: 0.5730

Iter 5802, loss: 0.5623 Iter 5803, loss: 0.5859 Iter 5804, loss: 0.6225 Iter 5805, loss: 0.5788 Iter 5806, loss: 0.5869 Iter 5807, loss: 0.5810 Iter 5808, loss: 0.5529 Iter 5809, loss: 0.5956 Iter 5810, loss: 0.5859 Iter 5811, loss: 0.5982 Iter 5812, loss: 0.6015 Iter 5813, loss: 0.5359 Iter 5814, loss: 0.5398 Iter 5815, loss: 0.5978 Iter 5816, loss: 0.5686 Iter 5817, loss: 0.5985 Iter 5818, loss: 0.5742 Iter 5819, loss: 0.5224 Iter 5820, loss: 0.5742 Iter 5821, loss: 0.5619 Iter 5822, loss: 0.6060 Iter 5823, loss: 0.6166 Iter 5824, loss: 0.5842 Iter 5825, loss: 0.5191 Iter 5826, loss: 0.5639 Iter 5827, loss: 0.5633 Iter 5828, loss: 0.5679 Iter 5829, loss: 0.5324 Iter 5830, loss: 0.5461 Iter 5831, loss: 0.5767 Iter 5832, loss: 0.5713 Iter 5833, loss: 0.5282 Iter 5834, loss: 0.5718 Iter 5835, loss: 0.5963 Iter 5836, loss: 0.6123 Iter 5837, loss: 0.5555 Iter 5838, loss: 0.5541 Iter 5839, loss: 0.5716 Iter 5840, loss: 0.6021 Iter 5841, loss: 0.5760 Iter 5842, loss: 0.5522 Iter 5843, loss: 0.5692

Iter 5844, loss: 0.5285 Iter 5845, loss: 0.6092 Iter 5846, loss: 0.5542 Iter 5847, loss: 0.5761 Iter 5848, loss: 0.5894 Iter 5849, loss: 0.5474 Iter 5850, loss: 0.5883 Iter 5851, loss: 0.5916 Iter 5852, loss: 0.5878 Iter 5853, loss: 0.5415 Iter 5854, loss: 0.6169 Iter 5855, loss: 0.5878 Iter 5856, loss: 0.5606 Iter 5857, loss: 0.5593 Iter 5858, loss: 0.5937 Iter 5859, loss: 0.5918 Iter 5860, loss: 0.5845 Iter 5861, loss: 0.5458 Iter 5862, loss: 0.5406 Iter 5863, loss: 0.5509 Iter 5864, loss: 0.5784 Iter 5865, loss: 0.5783 Iter 5866, loss: 0.5636 Iter 5867, loss: 0.5717 Iter 5868, loss: 0.6062 Iter 5869, loss: 0.5964 Iter 5870, loss: 0.6049 Iter 5871, loss: 0.5435 Iter 5872, loss: 0.5447 Iter 5873, loss: 0.6102 Iter 5874, loss: 0.5561 Iter 5875, loss: 0.6013 Iter 5876, loss: 0.5566 Iter 5877, loss: 0.5829 Iter 5878, loss: 0.5761 Iter 5879, loss: 0.6068 Iter 5880, loss: 0.6040 Iter 5881, loss: 0.5450 Iter 5882, loss: 0.5650 Iter 5883, loss: 0.5986 Iter 5884, loss: 0.6006 Iter 5885, loss: 0.6258

Iter 5886, loss: 0.5910 Iter 5887, loss: 0.5711 Iter 5888, loss: 0.5721 Iter 5889, loss: 0.6135 Iter 5890, loss: 0.5956 Iter 5891, loss: 0.5773 Iter 5892, loss: 0.5952 Iter 5893, loss: 0.5482 Iter 5894, loss: 0.5195 Iter 5895, loss: 0.5968 Iter 5896, loss: 0.5652 Iter 5897, loss: 0.5476 Iter 5898, loss: 0.5654 Iter 5899, loss: 0.5619 Iter 5900, loss: 0.5591 Iter 5901, loss: 0.5559 Iter 5902, loss: 0.5602 Iter 5903, loss: 0.6036 Iter 5904, loss: 0.5823 Iter 5905, loss: 0.5696 Iter 5906, loss: 0.5697 Iter 5907, loss: 0.5651 Iter 5908, loss: 0.5533 Iter 5909, loss: 0.5725 Iter 5910, loss: 0.5928 Iter 5911, loss: 0.5812 Iter 5912, loss: 0.5810 Iter 5913, loss: 0.5305 Iter 5914, loss: 0.6174 Iter 5915, loss: 0.5315 Iter 5916, loss: 0.5885 Iter 5917, loss: 0.5865 Iter 5918, loss: 0.5349 Iter 5919, loss: 0.5787 Iter 5920, loss: 0.5952 Iter 5921, loss: 0.5705 Iter 5922, loss: 0.5655 Iter 5923, loss: 0.6131 Iter 5924, loss: 0.6293 Iter 5925, loss: 0.5814 Iter 5926, loss: 0.5583

Iter 5927, loss: 0.5823 Iter 5928, loss: 0.5573 Iter 5929, loss: 0.5764 Iter 5930, loss: 0.5917 Iter 5931, loss: 0.5484 Iter 5932, loss: 0.5570 Iter 5933, loss: 0.5701 Iter 5934, loss: 0.5417 Iter 5935, loss: 0.5652 Iter 5936, loss: 0.5723 Iter 5937, loss: 0.5666 Iter 5938, loss: 0.5843 Iter 5939, loss: 0.5718 Iter 5940, loss: 0.5330 Iter 5941, loss: 0.5269 Iter 5942, loss: 0.5272 Iter 5943, loss: 0.5459 Iter 5944, loss: 0.5754 Iter 5945, loss: 0.5818 Iter 5946, loss: 0.5342 Iter 5947, loss: 0.5680 Iter 5948, loss: 0.6102 Iter 5949, loss: 0.5405 Iter 5950, loss: 0.5725 Iter 5951, loss: 0.5634 Iter 5952, loss: 0.6105 Iter 5953, loss: 0.6227 Iter 5954, loss: 0.5575 Iter 5955, loss: 0.5855 Iter 5956, loss: 0.6040 Iter 5957, loss: 0.5959 Iter 5958, loss: 0.5477 Iter 5959, loss: 0.5695 Iter 5960, loss: 0.5691 Iter 5961, loss: 0.5514 Iter 5962, loss: 0.5837 Iter 5963, loss: 0.5516 Iter 5964, loss: 0.5914 Iter 5965, loss: 0.5606 Iter 5966, loss: 0.5712 Iter 5967, loss: 0.5840 Iter 5968, loss: 0.5551

Iter 5969, loss: 0.5725 Iter 5970, loss: 0.5930 Iter 5971, loss: 0.5898 Iter 5972, loss: 0.5640 Iter 5973, loss: 0.5705 Iter 5974, loss: 0.5548 Iter 5975, loss: 0.5974 Iter 5976, loss: 0.5608 Iter 5977, loss: 0.5494 Iter 5978, loss: 0.5335 Iter 5979, loss: 0.5445 Iter 5980, loss: 0.5999 Iter 5981, loss: 0.5559 Iter 5982, loss: 0.5569 Iter 5983, loss: 0.5783 Iter 5984, loss: 0.5678 Iter 5985, loss: 0.6477 Iter 5986, loss: 0.5753 Iter 5987, loss: 0.5904 Iter 5988, loss: 0.5552 Iter 5989, loss: 0.5507 Iter 5990, loss: 0.5855 Iter 5991, loss: 0.5621 Iter 5992, loss: 0.5672 Iter 5993, loss: 0.5929 Iter 5994, loss: 0.5698 Iter 5995, loss: 0.5702 Iter 5996, loss: 0.5711 Iter 5997, loss: 0.5991 Iter 5998, loss: 0.5715 Iter 5999, loss: 0.5396 Iter 6000, loss: 0.5485 Iter 6001, loss: 0.5978 Iter 6002, loss: 0.5450 Iter 6003, loss: 0.5902 Iter 6004, loss: 0.5754 Iter 6005, loss: 0.5792 Iter 6006, loss: 0.5865 Iter 6007, loss: 0.5409 Iter 6008, loss: 0.6184 Iter 6009, loss: 0.5878

Iter 6010, loss: 0.5650 Iter 6011, loss: 0.6009 Iter 6012, loss: 0.6040 Iter 6013, loss: 0.5396 Iter 6014, loss: 0.5497 Iter 6015, loss: 0.5635 Iter 6016, loss: 0.5548 Iter 6017, loss: 0.5278 Iter 6018, loss: 0.5725 Iter 6019, loss: 0.5722 Iter 6020, loss: 0.5754 Iter 6021, loss: 0.5683 Iter 6022, loss: 0.5835 Iter 6023, loss: 0.5459 Iter 6024, loss: 0.5945 Iter 6025, loss: 0.5565 Iter 6026, loss: 0.6359 Iter 6027, loss: 0.5660 Iter 6028, loss: 0.5865 Iter 6029, loss: 0.5626 Iter 6030, loss: 0.5820 Iter 6031, loss: 0.5728 Iter 6032, loss: 0.5484 Iter 6033, loss: 0.5883 Iter 6034, loss: 0.5671 Iter 6035, loss: 0.5169 Iter 6036, loss: 0.5313 Iter 6037, loss: 0.5725 Iter 6038, loss: 0.5940 Iter 6039, loss: 0.5707 Iter 6040, loss: 0.5520 Iter 6041, loss: 0.6351 Iter 6042, loss: 0.5640 Iter 6043, loss: 0.5665 Iter 6044, loss: 0.6234 Iter 6045, loss: 0.5796 Iter 6046, loss: 0.5955 Iter 6047, loss: 0.5610 Iter 6048, loss: 0.5592 Iter 6049, loss: 0.5207 Iter 6050, loss: 0.5520

Iter 6051, loss: 0.6084 Iter 6052, loss: 0.5552 Iter 6053, loss: 0.5764 Iter 6054, loss: 0.5624 Iter 6055, loss: 0.5659 Iter 6056, loss: 0.6046 Iter 6057, loss: 0.5965 Iter 6058, loss: 0.5668 Iter 6059, loss: 0.6230 Iter 6060, loss: 0.5394 Iter 6061, loss: 0.5261 Iter 6062, loss: 0.5356 Iter 6063, loss: 0.5422 Iter 6064, loss: 0.5910 Iter 6065, loss: 0.5778 Iter 6066, loss: 0.6148 Iter 6067, loss: 0.5425 Iter 6068, loss: 0.5733 Iter 6069, loss: 0.5852 Iter 6070, loss: 0.5204 Iter 6071, loss: 0.5658 Iter 6072, loss: 0.5980 Iter 6073, loss: 0.5926 Iter 6074, loss: 0.5647 Iter 6075, loss: 0.5604 Iter 6076, loss: 0.5446 Iter 6077, loss: 0.5525 Iter 6078, loss: 0.5754 Iter 6079, loss: 0.5777 Iter 6080, loss: 0.5682 Iter 6081, loss: 0.5499 Iter 6082, loss: 0.5419 Iter 6083, loss: 0.5601 Iter 6084, loss: 0.5303 Iter 6085, loss: 0.5519 Iter 6086, loss: 0.6088 Iter 6087, loss: 0.5949 Iter 6088, loss: 0.5629 Iter 6089, loss: 0.5877 Iter 6090, loss: 0.5345 Iter 6091, loss: 0.5570 Iter 6092, loss: 0.5823

Iter 6093, loss: 0.5914 Iter 6094, loss: 0.5653 Iter 6095, loss: 0.5654 Iter 6096, loss: 0.5918 Iter 6097, loss: 0.5983 Iter 6098, loss: 0.5946 Iter 6099, loss: 0.5809 Iter 6100, loss: 0.5629 Iter 6101, loss: 0.6078 Iter 6102, loss: 0.5801 Iter 6103, loss: 0.5696 Iter 6104, loss: 0.5734 Iter 6105, loss: 0.5419 Iter 6106, loss: 0.5904 Iter 6107, loss: 0.5706 Iter 6108, loss: 0.5435 Iter 6109, loss: 0.5440 Iter 6110, loss: 0.5812 Iter 6111, loss: 0.5681 Iter 6112, loss: 0.6019 Iter 6113, loss: 0.5723 Iter 6114, loss: 0.6000 Iter 6115, loss: 0.5457 Iter 6116, loss: 0.6011 Iter 6117, loss: 0.5815 Iter 6118, loss: 0.5863 Iter 6119, loss: 0.5776 Iter 6120, loss: 0.5641 Iter 6121, loss: 0.5393 Iter 6122, loss: 0.5334 Iter 6123, loss: 0.6262 Iter 6124, loss: 0.5871 Iter 6125, loss: 0.5874 Iter 6126, loss: 0.5559 Iter 6127, loss: 0.5532 Iter 6128, loss: 0.5928 Iter 6129, loss: 0.5332 Iter 6130, loss: 0.5575 Iter 6131, loss: 0.5954 Iter 6132, loss: 0.5689 Iter 6133, loss: 0.5332

Iter 6134, loss: 0.5776 Iter 6135, loss: 0.5550 Iter 6136, loss: 0.5903 Iter 6137, loss: 0.5534 Iter 6138, loss: 0.5209 Iter 6139, loss: 0.5765 Iter 6140, loss: 0.5880 Iter 6141, loss: 0.5522 Iter 6142, loss: 0.5756 Iter 6143, loss: 0.5881 Iter 6144, loss: 0.5583 Iter 6145, loss: 0.5553 Iter 6146, loss: 0.6044 Iter 6147, loss: 0.5621 Iter 6148, loss: 0.5433 Iter 6149, loss: 0.5716 Iter 6150, loss: 0.5482 Iter 6151, loss: 0.6235 Iter 6152, loss: 0.5618 Iter 6153, loss: 0.5892 Iter 6154, loss: 0.5360 Iter 6155, loss: 0.5926 Iter 6156, loss: 0.5629 Iter 6157, loss: 0.5842 Iter 6158, loss: 0.6310 Iter 6159, loss: 0.5841 Iter 6160, loss: 0.6244 Iter 6161, loss: 0.5588 Iter 6162, loss: 0.5751 Iter 6163, loss: 0.5647 Iter 6164, loss: 0.5607 Iter 6165, loss: 0.5639 Iter 6166, loss: 0.5575 Iter 6167, loss: 0.5852 Iter 6168, loss: 0.5903 Iter 6169, loss: 0.5983 Iter 6170, loss: 0.5618 Iter 6171, loss: 0.5477 Iter 6172, loss: 0.5871 Iter 6173, loss: 0.5729 Iter 6174, loss: 0.5876 Iter 6175, loss: 0.5869

Iter 6176, loss: 0.5897 Iter 6177, loss: 0.5886 Iter 6178, loss: 0.5841 Iter 6179, loss: 0.5713 Iter 6180, loss: 0.5659 Iter 6181, loss: 0.5635 Iter 6182, loss: 0.6147 Iter 6183, loss: 0.6168 Iter 6184, loss: 0.5018 Iter 6185, loss: 0.5476 Iter 6186, loss: 0.5826 Iter 6187, loss: 0.5681 Iter 6188, loss: 0.5910 Iter 6189, loss: 0.5418 Iter 6190, loss: 0.5639 Iter 6191, loss: 0.5760 Iter 6192, loss: 0.5628 Iter 6193, loss: 0.6183 Iter 6194, loss: 0.5898 Iter 6195, loss: 0.5511 Iter 6196, loss: 0.5791 Iter 6197, loss: 0.5987 Iter 6198, loss: 0.6314 Iter 6199, loss: 0.5450 Iter 6200, loss: 0.5792 Iter 6201, loss: 0.5605 Iter 6202, loss: 0.5587 Iter 6203, loss: 0.5911 Iter 6204, loss: 0.5676 Iter 6205, loss: 0.5424 Iter 6206, loss: 0.5915 Iter 6207, loss: 0.5794 Iter 6208, loss: 0.6022 Iter 6209, loss: 0.5663 Iter 6210, loss: 0.5326 Iter 6211, loss: 0.5542 Iter 6212, loss: 0.6145 Iter 6213, loss: 0.5620 Iter 6214, loss: 0.5443 Iter 6215, loss: 0.5066 Iter 6216, loss: 0.5204

Iter 6217, loss: 0.5603 Iter 6218, loss: 0.5775 Iter 6219, loss: 0.5750 Iter 6220, loss: 0.5923 Iter 6221, loss: 0.6009 Iter 6222, loss: 0.5556 Iter 6223, loss: 0.5997 Iter 6224, loss: 0.5520 Iter 6225, loss: 0.5868 Iter 6226, loss: 0.5476 Iter 6227, loss: 0.5579 Iter 6228, loss: 0.5651 Iter 6229, loss: 0.5375 Iter 6230, loss: 0.5583 Iter 6231, loss: 0.5576 Iter 6232, loss: 0.5744 Iter 6233, loss: 0.5494 Iter 6234, loss: 0.6167 Iter 6235, loss: 0.5658 Iter 6236, loss: 0.5365 Iter 6237, loss: 0.5456 Iter 6238, loss: 0.5479 Iter 6239, loss: 0.5487 Iter 6240, loss: 0.5448 Iter 6241, loss: 0.5506 Iter 6242, loss: 0.5833 Iter 6243, loss: 0.5642 Iter 6244, loss: 0.5888 Iter 6245, loss: 0.5594 Iter 6246, loss: 0.5785 Iter 6247, loss: 0.5556 Iter 6248, loss: 0.5598 Iter 6249, loss: 0.5974 Iter 6250, loss: 0.5798 Iter 6251, loss: 0.5437 Iter 6252, loss: 0.5468 Iter 6253, loss: 0.5457 Iter 6254, loss: 0.5799 Iter 6255, loss: 0.5840 Iter 6256, loss: 0.5488 Iter 6257, loss: 0.5650

Iter 6258, loss: 0.5825 Iter 6259, loss: 0.5601 Iter 6260, loss: 0.6187 Iter 6261, loss: 0.5364 Iter 6262, loss: 0.5701 Iter 6263, loss: 0.5612 Iter 6264, loss: 0.5826 Iter 6265, loss: 0.5374 Iter 6266, loss: 0.5930 Iter 6267, loss: 0.5917 Iter 6268, loss: 0.5201 Iter 6269, loss: 0.5776 Iter 6270, loss: 0.6052 Iter 6271, loss: 0.5671 Iter 6272, loss: 0.5302 Iter 6273, loss: 0.5717 Iter 6274, loss: 0.5556 Iter 6275, loss: 0.5589 Iter 6276, loss: 0.5927 Iter 6277, loss: 0.5660 Iter 6278, loss: 0.5705 Iter 6279, loss: 0.5587 Iter 6280, loss: 0.5424 Iter 6281, loss: 0.5678 Iter 6282, loss: 0.5383 Iter 6283, loss: 0.5530 Iter 6284, loss: 0.5728 Iter 6285, loss: 0.5754 Iter 6286, loss: 0.5431 Iter 6287, loss: 0.5602 Iter 6288, loss: 0.5609 Iter 6289, loss: 0.5829 Iter 6290, loss: 0.5565 Iter 6291, loss: 0.5731 Iter 6292, loss: 0.6126 Iter 6293, loss: 0.5614 Iter 6294, loss: 0.6236 Iter 6295, loss: 0.5504 Iter 6296, loss: 0.5902 Iter 6297, loss: 0.5519 Iter 6298, loss: 0.5592 Iter 6299, loss: 0.5692

Iter 6300, loss: 0.5621 Iter 6301, loss: 0.5344 Iter 6302, loss: 0.5751 Iter 6303, loss: 0.5380 Iter 6304, loss: 0.5741 Iter 6305, loss: 0.5407 Iter 6306, loss: 0.5728 Iter 6307, loss: 0.5768 Iter 6308, loss: 0.5437 Iter 6309, loss: 0.5660 Iter 6310, loss: 0.5764 Iter 6311, loss: 0.5555 Iter 6312, loss: 0.5712 Iter 6313, loss: 0.5835 Iter 6314, loss: 0.5693 Iter 6315, loss: 0.5766 Iter 6316, loss: 0.5999 Iter 6317, loss: 0.5509 Iter 6318, loss: 0.5370 Iter 6319, loss: 0.5764 Iter 6320, loss: 0.5607 Iter 6321, loss: 0.5570 Iter 6322, loss: 0.5777 Iter 6323, loss: 0.5394 Iter 6324, loss: 0.5603 Iter 6325, loss: 0.5745 Iter 6326, loss: 0.5714 Iter 6327, loss: 0.5314 Iter 6328, loss: 0.5240 Iter 6329, loss: 0.6117 Iter 6330, loss: 0.5783 Iter 6331, loss: 0.5901 Iter 6332, loss: 0.6187 Iter 6333, loss: 0.5949 Iter 6334, loss: 0.5110 Iter 6335, loss: 0.5585 Iter 6336, loss: 0.5734 Iter 6337, loss: 0.5529 Iter 6338, loss: 0.5583 Iter 6339, loss: 0.5730 Iter 6340, loss: 0.6194

Iter 6341, loss: 0.6060 Iter 6342, loss: 0.6079 Iter 6343, loss: 0.5793 Iter 6344, loss: 0.6105 Iter 6345, loss: 0.5649 Iter 6346, loss: 0.5845 Iter 6347, loss: 0.5656 Iter 6348, loss: 0.5682 Iter 6349, loss: 0.5499 Iter 6350, loss: 0.5737 Iter 6351, loss: 0.5646 Iter 6352, loss: 0.5592 Iter 6353, loss: 0.5880 Iter 6354, loss: 0.6196 Iter 6355, loss: 0.5569 Iter 6356, loss: 0.5741 Iter 6357, loss: 0.5423 Iter 6358, loss: 0.5695 Iter 6359, loss: 0.5332 Iter 6360, loss: 0.5130 Iter 6361, loss: 0.6237 Iter 6362, loss: 0.5880 Iter 6363, loss: 0.5573 Iter 6364, loss: 0.5427 Iter 6365, loss: 0.6101 Iter 6366, loss: 0.5927 Iter 6367, loss: 0.5917 Iter 6368, loss: 0.5727 Iter 6369, loss: 0.5826 Iter 6370, loss: 0.5477 Iter 6371, loss: 0.5604 Iter 6372, loss: 0.5588 Iter 6373, loss: 0.5524 Iter 6374, loss: 0.5766 Iter 6375, loss: 0.5488 Iter 6376, loss: 0.5532 Iter 6377, loss: 0.5826 Iter 6378, loss: 0.5976 Iter 6379, loss: 0.6187 Iter 6380, loss: 0.5820 Iter 6381, loss: 0.5996 Iter 6382, loss: 0.5542

Iter 6383, loss: 0.5518 Iter 6384, loss: 0.5748 Iter 6385, loss: 0.5630 Iter 6386, loss: 0.5321 Iter 6387, loss: 0.5992 Iter 6388, loss: 0.6024 Iter 6389, loss: 0.5468 Iter 6390, loss: 0.5506 Iter 6391, loss: 0.5530 Iter 6392, loss: 0.5763 Iter 6393, loss: 0.5636 Iter 6394, loss: 0.5404 Iter 6395, loss: 0.5155 Iter 6396, loss: 0.5506 Iter 6397, loss: 0.6032 Iter 6398, loss: 0.5905 Iter 6399, loss: 0.5592 Iter 6400, loss: 0.5732 Iter 6401, loss: 0.5802 Iter 6402, loss: 0.5312 Iter 6403, loss: 0.5600 Iter 6404, loss: 0.5571 Iter 6405, loss: 0.5884 Iter 6406, loss: 0.5470 Iter 6407, loss: 0.5630 Iter 6408, loss: 0.5842 Iter 6409, loss: 0.5732 Iter 6410, loss: 0.5695 Iter 6411, loss: 0.5383 Iter 6412, loss: 0.5859 Iter 6413, loss: 0.5522 Iter 6414, loss: 0.5555 Iter 6415, loss: 0.5548 Iter 6416, loss: 0.5368 Iter 6417, loss: 0.5452 Iter 6418, loss: 0.5059 Iter 6419, loss: 0.5399 Iter 6420, loss: 0.5792 Iter 6421, loss: 0.5359 Iter 6422, loss: 0.5602 Iter 6423, loss: 0.5665 Iter 6424, loss: 0.5535

Iter 6425, loss: 0.5788 Iter 6426, loss: 0.5457 Iter 6427, loss: 0.5854 Iter 6428, loss: 0.6052 Iter 6429, loss: 0.5949 Iter 6430, loss: 0.5878 Iter 6431, loss: 0.5761 Iter 6432, loss: 0.5588 Iter 6433, loss: 0.6151 Iter 6434, loss: 0.5015 Iter 6435, loss: 0.5390 Iter 6436, loss: 0.5918 Iter 6437, loss: 0.5755 Iter 6438, loss: 0.6159 Iter 6439, loss: 0.5766 Iter 6440, loss: 0.5153 Iter 6441, loss: 0.6018 Iter 6442, loss: 0.6097 Iter 6443, loss: 0.5711 Iter 6444, loss: 0.5375 Iter 6445, loss: 0.5653 Iter 6446, loss: 0.5921 Iter 6447, loss: 0.5456 Iter 6448, loss: 0.5228 Iter 6449, loss: 0.5759 Iter 6450, loss: 0.5641 Iter 6451, loss: 0.5666 Iter 6452, loss: 0.5229 Iter 6453, loss: 0.5405 Iter 6454, loss: 0.5617 Iter 6455, loss: 0.5347 Iter 6456, loss: 0.5546 Iter 6457, loss: 0.5456 Iter 6458, loss: 0.5952 Iter 6459, loss: 0.5675 Iter 6460, loss: 0.5238 Iter 6461, loss: 0.5622 Iter 6462, loss: 0.5284 Iter 6463, loss: 0.5145 Iter 6464, loss: 0.5661 Iter 6465, loss: 0.5847

Iter 6466, loss: 0.5474 Iter 6467, loss: 0.5272 Iter 6468, loss: 0.6144 Iter 6469, loss: 0.6014 Iter 6470, loss: 0.5503 Iter 6471, loss: 0.5236 Iter 6472, loss: 0.5492 Iter 6473, loss: 0.5977 Iter 6474, loss: 0.5870 Iter 6475, loss: 0.5996 Iter 6476, loss: 0.5783 Iter 6477, loss: 0.5747 Iter 6478, loss: 0.5574 Iter 6479, loss: 0.5455 Iter 6480, loss: 0.5774 Iter 6481, loss: 0.5503 Iter 6482, loss: 0.5578 Iter 6483, loss: 0.5481 Iter 6484, loss: 0.5728 Iter 6485, loss: 0.5876 Iter 6486, loss: 0.5133 Iter 6487, loss: 0.6046 Iter 6488, loss: 0.5764 Iter 6489, loss: 0.5388 Iter 6490, loss: 0.5786 Iter 6491, loss: 0.5653 Iter 6492, loss: 0.5455 Iter 6493, loss: 0.5640 Iter 6494, loss: 0.5801 Iter 6495, loss: 0.5594 Iter 6496, loss: 0.5753 Iter 6497, loss: 0.5459 Iter 6498, loss: 0.5654 Iter 6499, loss: 0.5490 Iter 6500, loss: 0.5874 Iter 6501, loss: 0.5439 Iter 6502, loss: 0.5764 Iter 6503, loss: 0.5373 Iter 6504, loss: 0.5537 Iter 6505, loss: 0.5838 Iter 6506, loss: 0.6190 Iter 6507, loss: 0.5624

Iter 6508, loss: 0.5343 Iter 6509, loss: 0.5801 Iter 6510, loss: 0.5993 Iter 6511, loss: 0.5558 Iter 6512, loss: 0.5327 Iter 6513, loss: 0.5428 Iter 6514, loss: 0.5610 Iter 6515, loss: 0.5657 Iter 6516, loss: 0.5325 Iter 6517, loss: 0.5591 Iter 6518, loss: 0.5546 Iter 6519, loss: 0.5506 Iter 6520, loss: 0.5968 Iter 6521, loss: 0.6294 Iter 6522, loss: 0.5763 Iter 6523, loss: 0.5588 Iter 6524, loss: 0.5603 Iter 6525, loss: 0.5794 Iter 6526, loss: 0.5247 Iter 6527, loss: 0.5989 Iter 6528, loss: 0.5932 Iter 6529, loss: 0.5641 Iter 6530, loss: 0.5679 Iter 6531, loss: 0.5126 Iter 6532, loss: 0.5610 Iter 6533, loss: 0.5267 Iter 6534, loss: 0.5553 Iter 6535, loss: 0.5670 Iter 6536, loss: 0.5579 Iter 6537, loss: 0.5460 Iter 6538, loss: 0.5560 Iter 6539, loss: 0.5473 Iter 6540, loss: 0.5426 Iter 6541, loss: 0.5670 Iter 6542, loss: 0.5547 Iter 6543, loss: 0.5441 Iter 6544, loss: 0.5690 Iter 6545, loss: 0.5632 Iter 6546, loss: 0.5306 Iter 6547, loss: 0.5675 Iter 6548, loss: 0.5956 Iter 6549, loss: 0.5936

Iter 6550, loss: 0.5694 Iter 6551, loss: 0.5505 Iter 6552, loss: 0.5619 Iter 6553, loss: 0.5690 Iter 6554, loss: 0.5610 Iter 6555, loss: 0.5770 Iter 6556, loss: 0.5557 Iter 6557, loss: 0.5486 Iter 6558, loss: 0.5142 Iter 6559, loss: 0.5927 Iter 6560, loss: 0.5166 Iter 6561, loss: 0.5806 Iter 6562, loss: 0.5688 Iter 6563, loss: 0.5524 Iter 6564, loss: 0.5880 Iter 6565, loss: 0.6298 Iter 6566, loss: 0.5777 Iter 6567, loss: 0.5677 Iter 6568, loss: 0.5382 Iter 6569, loss: 0.5347 Iter 6570, loss: 0.5217 Iter 6571, loss: 0.5697 Iter 6572, loss: 0.5423 Iter 6573, loss: 0.5797 Iter 6574, loss: 0.5731 Iter 6575, loss: 0.5655 Iter 6576, loss: 0.5754 Iter 6577, loss: 0.5612 Iter 6578, loss: 0.5661 Iter 6579, loss: 0.5406 Iter 6580, loss: 0.5300 Iter 6581, loss: 0.5518 Iter 6582, loss: 0.5546 Iter 6583, loss: 0.5343 Iter 6584, loss: 0.5627 Iter 6585, loss: 0.5196 Iter 6586, loss: 0.6066 Iter 6587, loss: 0.5675 Iter 6588, loss: 0.5937 Iter 6589, loss: 0.5370 Iter 6590, loss: 0.5377 Iter 6591, loss: 0.5330

Iter 6592, loss: 0.5511 Iter 6593, loss: 0.5472 Iter 6594, loss: 0.5529 Iter 6595, loss: 0.5806 Iter 6596, loss: 0.5778 Iter 6597, loss: 0.5424 Iter 6598, loss: 0.5850 Iter 6599, loss: 0.5300 Iter 6600, loss: 0.5440 Iter 6601, loss: 0.5750 Iter 6602, loss: 0.6185 Iter 6603, loss: 0.5968 Iter 6604, loss: 0.5897 Iter 6605, loss: 0.5875 Iter 6606, loss: 0.5357 Iter 6607, loss: 0.5799 Iter 6608, loss: 0.5780 Iter 6609, loss: 0.5817 Iter 6610, loss: 0.5817 Iter 6611, loss: 0.5989 Iter 6612, loss: 0.5575 Iter 6613, loss: 0.5473 Iter 6614, loss: 0.5274 Iter 6615, loss: 0.5428 Iter 6616, loss: 0.5398 Iter 6617, loss: 0.5482 Iter 6618, loss: 0.5796 Iter 6619, loss: 0.6165 Iter 6620, loss: 0.5189 Iter 6621, loss: 0.5919 Iter 6622, loss: 0.5200 Iter 6623, loss: 0.5770 Iter 6624, loss: 0.5663 Iter 6625, loss: 0.5872 Iter 6626, loss: 0.5752 Iter 6627, loss: 0.5743 Iter 6628, loss: 0.5279 Iter 6629, loss: 0.5718 Iter 6630, loss: 0.5588 Iter 6631, loss: 0.5791 Iter 6632, loss: 0.5148 Iter 6633, loss: 0.5687

Iter 6634, loss: 0.5316 Iter 6635, loss: 0.5680 Iter 6636, loss: 0.5479 Iter 6637, loss: 0.6080 Iter 6638, loss: 0.5506 Iter 6639, loss: 0.5450 Iter 6640, loss: 0.5371 Iter 6641, loss: 0.5926 Iter 6642, loss: 0.5392 Iter 6643, loss: 0.5411 Iter 6644, loss: 0.5844 Iter 6645, loss: 0.5533 Iter 6646, loss: 0.5933 Iter 6647, loss: 0.5365 Iter 6648, loss: 0.5429 Iter 6649, loss: 0.5390 Iter 6650, loss: 0.5862 Iter 6651, loss: 0.6123 Iter 6652, loss: 0.6113 Iter 6653, loss: 0.5636 Iter 6654, loss: 0.5972 Iter 6655, loss: 0.5680 Iter 6656, loss: 0.5624 Iter 6657, loss: 0.5764 Iter 6658, loss: 0.5631 Iter 6659, loss: 0.5560 Iter 6660, loss: 0.5793 Iter 6661, loss: 0.5738 Iter 6662, loss: 0.5722 Iter 6663, loss: 0.5779 Iter 6664, loss: 0.5843 Iter 6665, loss: 0.5411 Iter 6666, loss: 0.5782 Iter 6667, loss: 0.5842 Iter 6668, loss: 0.5282 Iter 6669, loss: 0.5828 Iter 6670, loss: 0.5279 Iter 6671, loss: 0.5953 Iter 6672, loss: 0.5787 Iter 6673, loss: 0.5573 Iter 6674, loss: 0.5732

Iter 6675, loss: 0.5278 Iter 6676, loss: 0.5875 Iter 6677, loss: 0.5179 Iter 6678, loss: 0.5564 Iter 6679, loss: 0.5583 Iter 6680, loss: 0.5568 Iter 6681, loss: 0.5835 Iter 6682, loss: 0.5710 Iter 6683, loss: 0.5793 Iter 6684, loss: 0.5918 Iter 6685, loss: 0.5955 Iter 6686, loss: 0.5720 Iter 6687, loss: 0.5417 Iter 6688, loss: 0.5784 Iter 6689, loss: 0.5711 Iter 6690, loss: 0.6115 Iter 6691, loss: 0.5843 Iter 6692, loss: 0.5295 Iter 6693, loss: 0.5516 Iter 6694, loss: 0.5905 Iter 6695, loss: 0.5550 Iter 6696, loss: 0.5601 Iter 6697, loss: 0.4988 Iter 6698, loss: 0.5860 Iter 6699, loss: 0.5935 Iter 6700, loss: 0.5440 Iter 6701, loss: 0.5490 Iter 6702, loss: 0.5740 Iter 6703, loss: 0.5645 Iter 6704, loss: 0.5576 Iter 6705, loss: 0.5853 Iter 6706, loss: 0.5872 Iter 6707, loss: 0.5425 Iter 6708, loss: 0.5368 Iter 6709, loss: 0.5341 Iter 6710, loss: 0.5149 Iter 6711, loss: 0.5654 Iter 6712, loss: 0.5444 Iter 6713, loss: 0.5608 Iter 6714, loss: 0.5717 Iter 6715, loss: 0.5412

Iter 6716, loss: 0.5287 Iter 6717, loss: 0.5481 Iter 6718, loss: 0.5515 Iter 6719, loss: 0.5455 Iter 6720, loss: 0.5890 Iter 6721, loss: 0.5573 Iter 6722, loss: 0.5952 Iter 6723, loss: 0.6266 Iter 6724, loss: 0.5648 Iter 6725, loss: 0.5928 Iter 6726, loss: 0.5478 Iter 6727, loss: 0.5638 Iter 6728, loss: 0.5763 Iter 6729, loss: 0.5443 Iter 6730, loss: 0.5698 Iter 6731, loss: 0.5765 Iter 6732, loss: 0.5600 Iter 6733, loss: 0.5707 Iter 6734, loss: 0.5710 Iter 6735, loss: 0.5612 Iter 6736, loss: 0.5568 Iter 6737, loss: 0.5113 Iter 6738, loss: 0.5577 Iter 6739, loss: 0.5502 Iter 6740, loss: 0.5841 Iter 6741, loss: 0.5642 Iter 6742, loss: 0.5585 Iter 6743, loss: 0.5854 Iter 6744, loss: 0.5742 Iter 6745, loss: 0.5854 Iter 6746, loss: 0.5272 Iter 6747, loss: 0.5284 Iter 6748, loss: 0.5608 Iter 6749, loss: 0.5852 Iter 6750, loss: 0.5179 Iter 6751, loss: 0.5663 Iter 6752, loss: 0.5663 Iter 6753, loss: 0.5828 Iter 6754, loss: 0.5586 Iter 6755, loss: 0.5427 Iter 6756, loss: 0.6181

Iter 6757, loss: 0.5473 Iter 6758, loss: 0.5641 Iter 6759, loss: 0.5016 Iter 6760, loss: 0.6193 Iter 6761, loss: 0.5580 Iter 6762, loss: 0.5663 Iter 6763, loss: 0.5681 Iter 6764, loss: 0.5361 Iter 6765, loss: 0.5629 Iter 6766, loss: 0.5207 Iter 6767, loss: 0.5679 Iter 6768, loss: 0.5737 Iter 6769, loss: 0.5784 Iter 6770, loss: 0.5272 Iter 6771, loss: 0.6005 Iter 6772, loss: 0.5440 Iter 6773, loss: 0.5559 Iter 6774, loss: 0.6200 Iter 6775, loss: 0.5062 Iter 6776, loss: 0.5669 Iter 6777, loss: 0.5579 Iter 6778, loss: 0.5316 Iter 6779, loss: 0.5469 Iter 6780, loss: 0.5818 Iter 6781, loss: 0.5881 Iter 6782, loss: 0.5199 Iter 6783, loss: 0.5518 Iter 6784, loss: 0.5915 Iter 6785, loss: 0.5409 Iter 6786, loss: 0.5979 Iter 6787, loss: 0.5825 Iter 6788, loss: 0.6019 Iter 6789, loss: 0.5379 Iter 6790, loss: 0.5486 Iter 6791, loss: 0.5680 Iter 6792, loss: 0.5359 Iter 6793, loss: 0.5493 Iter 6794, loss: 0.5386 Iter 6795, loss: 0.5900 Iter 6796, loss: 0.5774 Iter 6797, loss: 0.5320 Iter 6798, loss: 0.5083

Iter 6799, loss: 0.5551 Iter 6800, loss: 0.5964 Iter 6801, loss: 0.5563 Iter 6802, loss: 0.5838 Iter 6803, loss: 0.5616 Iter 6804, loss: 0.5665 Iter 6805, loss: 0.5798 Iter 6806, loss: 0.5986 Iter 6807, loss: 0.5643 Iter 6808, loss: 0.5394 Iter 6809, loss: 0.5371 Iter 6810, loss: 0.5694 Iter 6811, loss: 0.5629 Iter 6812, loss: 0.5561 Iter 6813, loss: 0.5651 Iter 6814, loss: 0.5699 Iter 6815, loss: 0.6431 Iter 6816, loss: 0.6090 Iter 6817, loss: 0.5338 Iter 6818, loss: 0.5996 Iter 6819, loss: 0.5847 Iter 6820, loss: 0.6072 Iter 6821, loss: 0.5074 Iter 6822, loss: 0.6058 Iter 6823, loss: 0.5624 Iter 6824, loss: 0.5706 Iter 6825, loss: 0.5737 Iter 6826, loss: 0.5489 Iter 6827, loss: 0.6063 Iter 6828, loss: 0.5626 Iter 6829, loss: 0.5586 Iter 6830, loss: 0.5423 Iter 6831, loss: 0.5773 Iter 6832, loss: 0.5401 Iter 6833, loss: 0.5687 Iter 6834, loss: 0.5139 Iter 6835, loss: 0.5422 Iter 6836, loss: 0.5710 Iter 6837, loss: 0.5867 Iter 6838, loss: 0.5336 Iter 6839, loss: 0.5569 Iter 6840, loss: 0.5109

Iter 6841, loss: 0.5669 Iter 6842, loss: 0.5615 Iter 6843, loss: 0.5489 Iter 6844, loss: 0.5559 Iter 6845, loss: 0.5290 Iter 6846, loss: 0.5319 Iter 6847, loss: 0.5563 Iter 6848, loss: 0.5064 Iter 6849, loss: 0.5671 Iter 6850, loss: 0.5989 Iter 6851, loss: 0.5534 Iter 6852, loss: 0.6403 Iter 6853, loss: 0.5535 Iter 6854, loss: 0.5367 Iter 6855, loss: 0.5446 Iter 6856, loss: 0.5641 Iter 6857, loss: 0.5650 Iter 6858, loss: 0.5284 Iter 6859, loss: 0.5907 Iter 6860, loss: 0.5342 Iter 6861, loss: 0.5565 Iter 6862, loss: 0.5952 Iter 6863, loss: 0.5635 Iter 6864, loss: 0.5651 Iter 6865, loss: 0.5636 Iter 6866, loss: 0.5665 Iter 6867, loss: 0.5526 Iter 6868, loss: 0.5668 Iter 6869, loss: 0.5592 Iter 6870, loss: 0.5615 Iter 6871, loss: 0.5895 Iter 6872, loss: 0.6104 Iter 6873, loss: 0.5665 Iter 6874, loss: 0.5412 Iter 6875, loss: 0.5654 Iter 6876, loss: 0.5467 Iter 6877, loss: 0.5253 Iter 6878, loss: 0.5776 Iter 6879, loss: 0.5610 Iter 6880, loss: 0.5644 Iter 6881, loss: 0.5725 Iter 6882, loss: 0.5855

Iter 6883, loss: 0.5764 Iter 6884, loss: 0.5564 Iter 6885, loss: 0.5749 Iter 6886, loss: 0.5910 Iter 6887, loss: 0.5400 Iter 6888, loss: 0.5709 Iter 6889, loss: 0.5612 Iter 6890, loss: 0.5565 Iter 6891, loss: 0.5471 Iter 6892, loss: 0.5063 Iter 6893, loss: 0.5471 Iter 6894, loss: 0.5888 Iter 6895, loss: 0.6021 Iter 6896, loss: 0.5075 Iter 6897, loss: 0.5239 Iter 6898, loss: 0.5630 Iter 6899, loss: 0.5431 Iter 6900, loss: 0.5485 Iter 6901, loss: 0.5669 Iter 6902, loss: 0.5893 Iter 6903, loss: 0.5755 Iter 6904, loss: 0.5792 Iter 6905, loss: 0.5939 Iter 6906, loss: 0.5784 Iter 6907, loss: 0.5538 Iter 6908, loss: 0.5790 Iter 6909, loss: 0.5780 Iter 6910, loss: 0.5732 Iter 6911, loss: 0.5505 Iter 6912, loss: 0.5885 Iter 6913, loss: 0.6054 Iter 6914, loss: 0.5608 Iter 6915, loss: 0.5461 Iter 6916, loss: 0.6119 Iter 6917, loss: 0.5766 Iter 6918, loss: 0.5658 Iter 6919, loss: 0.5097 Iter 6920, loss: 0.5404 Iter 6921, loss: 0.5594 Iter 6922, loss: 0.5913 Iter 6923, loss: 0.5754 Iter 6924, loss: 0.5768

Iter 6925, loss: 0.5312 Iter 6926, loss: 0.5703 Iter 6927, loss: 0.5106 Iter 6928, loss: 0.5670 Iter 6929, loss: 0.5875 Iter 6930, loss: 0.5877 Iter 6931, loss: 0.5309 Iter 6932, loss: 0.5568 Iter 6933, loss: 0.6073 Iter 6934, loss: 0.5363 Iter 6935, loss: 0.5558 Iter 6936, loss: 0.5457 Iter 6937, loss: 0.5643 Iter 6938, loss: 0.5583 Iter 6939, loss: 0.5764 Iter 6940, loss: 0.5724 Iter 6941, loss: 0.5390 Iter 6942, loss: 0.5315 Iter 6943, loss: 0.5709 Iter 6944, loss: 0.5438 Iter 6945, loss: 0.5416 Iter 6946, loss: 0.5487 Iter 6947, loss: 0.5246 Iter 6948, loss: 0.5357 Iter 6949, loss: 0.5929 Iter 6950, loss: 0.5557 Iter 6951, loss: 0.5235 Iter 6952, loss: 0.5606 Iter 6953, loss: 0.5556 Iter 6954, loss: 0.5698 Iter 6955, loss: 0.6162 Iter 6956, loss: 0.5543 Iter 6957, loss: 0.5608 Iter 6958, loss: 0.5234 Iter 6959, loss: 0.5943 Iter 6960, loss: 0.5991 Iter 6961, loss: 0.5367 Iter 6962, loss: 0.5844 Iter 6963, loss: 0.5895 Iter 6964, loss: 0.5357 Iter 6965, loss: 0.5740 Iter 6966, loss: 0.5583

Iter 6967, loss: 0.5775 Iter 6968, loss: 0.5127 Iter 6969, loss: 0.5660 Iter 6970, loss: 0.5464 Iter 6971, loss: 0.5710 Iter 6972, loss: 0.5581 Iter 6973, loss: 0.5650 Iter 6974, loss: 0.5581 Iter 6975, loss: 0.5371 Iter 6976, loss: 0.5690 Iter 6977, loss: 0.5566 Iter 6978, loss: 0.5508 Iter 6979, loss: 0.5513 Iter 6980, loss: 0.5908 Iter 6981, loss: 0.5734 Iter 6982, loss: 0.5570 Iter 6983, loss: 0.5743 Iter 6984, loss: 0.5490 Iter 6985, loss: 0.5863 Iter 6986, loss: 0.5339 Iter 6987, loss: 0.5309 Iter 6988, loss: 0.5551 Iter 6989, loss: 0.5654 Iter 6990, loss: 0.5622 Iter 6991, loss: 0.5368 Iter 6992, loss: 0.5425 Iter 6993, loss: 0.4998 Iter 6994, loss: 0.5572 Iter 6995, loss: 0.5227 Iter 6996, loss: 0.5007 Iter 6997, loss: 0.5157 Iter 6998, loss: 0.5631 Iter 6999, loss: 0.5723 Iter 7000, loss: 0.6288 Iter 7001, loss: 0.5290 Iter 7002, loss: 0.5875 Iter 7003, loss: 0.5836 Iter 7004, loss: 0.5584 Iter 7005, loss: 0.5607 Iter 7006, loss: 0.5705 Iter 7007, loss: 0.5371 Iter 7008, loss: 0.6216

Iter 7009, loss: 0.5746 Iter 7010, loss: 0.5712 Iter 7011, loss: 0.6103 Iter 7012, loss: 0.5317 Iter 7013, loss: 0.5704 Iter 7014, loss: 0.5381 Iter 7015, loss: 0.5704 Iter 7016, loss: 0.5819 Iter 7017, loss: 0.5878 Iter 7018, loss: 0.5707 Iter 7019, loss: 0.5762 Iter 7020, loss: 0.5646 Iter 7021, loss: 0.5770 Iter 7022, loss: 0.5179 Iter 7023, loss: 0.5506 Iter 7024, loss: 0.5125 Iter 7025, loss: 0.5533 Iter 7026, loss: 0.5796 Iter 7027, loss: 0.5783 Iter 7028, loss: 0.5168 Iter 7029, loss: 0.5665 Iter 7030, loss: 0.5580 Iter 7031, loss: 0.5617 Iter 7032, loss: 0.5474 Iter 7033, loss: 0.5603 Iter 7034, loss: 0.5284 Iter 7035, loss: 0.5677 Iter 7036, loss: 0.5306 Iter 7037, loss: 0.5641 Iter 7038, loss: 0.5621 Iter 7039, loss: 0.5296 Iter 7040, loss: 0.5417 Iter 7041, loss: 0.5917 Iter 7042, loss: 0.5324 Iter 7043, loss: 0.5830 Iter 7044, loss: 0.5365 Iter 7045, loss: 0.5500 Iter 7046, loss: 0.5772 Iter 7047, loss: 0.5689 Iter 7048, loss: 0.5552 Iter 7049, loss: 0.5525 Iter 7050, loss: 0.5380

Iter 7051, loss: 0.5612 Iter 7052, loss: 0.5433 Iter 7053, loss: 0.5745 Iter 7054, loss: 0.5606 Iter 7055, loss: 0.5975 Iter 7056, loss: 0.6129 Iter 7057, loss: 0.5291 Iter 7058, loss: 0.5504 Iter 7059, loss: 0.5688 Iter 7060, loss: 0.5495 Iter 7061, loss: 0.6007 Iter 7062, loss: 0.5427 Iter 7063, loss: 0.5274 Iter 7064, loss: 0.5093 Iter 7065, loss: 0.5629 Iter 7066, loss: 0.5714 Iter 7067, loss: 0.5238 Iter 7068, loss: 0.5745 Iter 7069, loss: 0.5742 Iter 7070, loss: 0.5421 Iter 7071, loss: 0.5409 Iter 7072, loss: 0.6035 Iter 7073, loss: 0.5256 Iter 7074, loss: 0.5817 Iter 7075, loss: 0.5271 Iter 7076, loss: 0.5552 Iter 7077, loss: 0.5699 Iter 7078, loss: 0.5150 Iter 7079, loss: 0.6148 Iter 7080, loss: 0.6086 Iter 7081, loss: 0.5483 Iter 7082, loss: 0.5375 Iter 7083, loss: 0.5451 Iter 7084, loss: 0.5628 Iter 7085, loss: 0.5336 Iter 7086, loss: 0.5715 Iter 7087, loss: 0.5593 Iter 7088, loss: 0.5362 Iter 7089, loss: 0.5309 Iter 7090, loss: 0.5194 Iter 7091, loss: 0.5519 Iter 7092, loss: 0.5320

Iter 7093, loss: 0.5706 Iter 7094, loss: 0.5695 Iter 7095, loss: 0.5439 Iter 7096, loss: 0.5403 Iter 7097, loss: 0.5200 Iter 7098, loss: 0.5478 Iter 7099, loss: 0.5542 Iter 7100, loss: 0.5346 Iter 7101, loss: 0.6357 Iter 7102, loss: 0.5112 Iter 7103, loss: 0.6229 Iter 7104, loss: 0.5569 Iter 7105, loss: 0.6072 Iter 7106, loss: 0.5344 Iter 7107, loss: 0.5449 Iter 7108, loss: 0.5133 Iter 7109, loss: 0.5407 Iter 7110, loss: 0.5386 Iter 7111, loss: 0.5320 Iter 7112, loss: 0.5655 Iter 7113, loss: 0.5700 Iter 7114, loss: 0.5942 Iter 7115, loss: 0.5476 Iter 7116, loss: 0.5602 Iter 7117, loss: 0.5509 Iter 7118, loss: 0.5964 Iter 7119, loss: 0.5425 Iter 7120, loss: 0.5648 Iter 7121, loss: 0.5492 Iter 7122, loss: 0.5832 Iter 7123, loss: 0.5789 Iter 7124, loss: 0.5666 Iter 7125, loss: 0.5633 Iter 7126, loss: 0.5243 Iter 7127, loss: 0.5154 Iter 7128, loss: 0.5419 Iter 7129, loss: 0.5727 Iter 7130, loss: 0.5731 Iter 7131, loss: 0.6048 Iter 7132, loss: 0.5827 Iter 7133, loss: 0.5941 Iter 7134, loss: 0.5292

Iter 7135, loss: 0.5520 Iter 7136, loss: 0.5370 Iter 7137, loss: 0.5501 Iter 7138, loss: 0.5444 Iter 7139, loss: 0.5412 Iter 7140, loss: 0.5703 Iter 7141, loss: 0.5840 Iter 7142, loss: 0.5419 Iter 7143, loss: 0.5946 Iter 7144, loss: 0.6182 Iter 7145, loss: 0.5487 Iter 7146, loss: 0.5381 Iter 7147, loss: 0.5339 Iter 7148, loss: 0.5250 Iter 7149, loss: 0.5481 Iter 7150, loss: 0.5477 Iter 7151, loss: 0.5651 Iter 7152, loss: 0.5279 Iter 7153, loss: 0.5987 Iter 7154, loss: 0.5390 Iter 7155, loss: 0.5926 Iter 7156, loss: 0.6008 Iter 7157, loss: 0.5583 Iter 7158, loss: 0.5577 Iter 7159, loss: 0.5982 Iter 7160, loss: 0.5983 Iter 7161, loss: 0.5584 Iter 7162, loss: 0.5314 Iter 7163, loss: 0.5596 Iter 7164, loss: 0.5686 Iter 7165, loss: 0.5740 Iter 7166, loss: 0.5472 Iter 7167, loss: 0.5581 Iter 7168, loss: 0.5628 Iter 7169, loss: 0.5290 Iter 7170, loss: 0.5233 Iter 7171, loss: 0.5676 Iter 7172, loss: 0.5089 Iter 7173, loss: 0.5385 Iter 7174, loss: 0.5644 Iter 7175, loss: 0.5519 Iter 7176, loss: 0.5973

Iter 7177, loss: 0.5486 Iter 7178, loss: 0.5739 Iter 7179, loss: 0.5526 Iter 7180, loss: 0.5527 Iter 7181, loss: 0.5738 Iter 7182, loss: 0.5361 Iter 7183, loss: 0.5022 Iter 7184, loss: 0.5795 Iter 7185, loss: 0.5877 Iter 7186, loss: 0.5584 Iter 7187, loss: 0.5745 Iter 7188, loss: 0.5563 Iter 7189, loss: 0.5202 Iter 7190, loss: 0.6093 Iter 7191, loss: 0.4843 Iter 7192, loss: 0.5781 Iter 7193, loss: 0.5278 Iter 7194, loss: 0.5364 Iter 7195, loss: 0.5959 Iter 7196, loss: 0.5745 Iter 7197, loss: 0.5181 Iter 7198, loss: 0.5261 Iter 7199, loss: 0.5660 Iter 7200, loss: 0.5413 Iter 7201, loss: 0.5407 Iter 7202, loss: 0.5631 Iter 7203, loss: 0.5687 Iter 7204, loss: 0.5561 Iter 7205, loss: 0.5048 Iter 7206, loss: 0.5782 Iter 7207, loss: 0.4837 Iter 7208, loss: 0.5776 Iter 7209, loss: 0.5702 Iter 7210, loss: 0.5527 Iter 7211, loss: 0.5579 Iter 7212, loss: 0.5884 Iter 7213, loss: 0.5870 Iter 7214, loss: 0.5806 Iter 7215, loss: 0.5584 Iter 7216, loss: 0.5276 Iter 7217, loss: 0.5192

Iter 7218, loss: 0.5608 Iter 7219, loss: 0.4989 Iter 7220, loss: 0.5349 Iter 7221, loss: 0.5222 Iter 7222, loss: 0.5652 Iter 7223, loss: 0.5643 Iter 7224, loss: 0.5160 Iter 7225, loss: 0.5722 Iter 7226, loss: 0.5447 Iter 7227, loss: 0.5495 Iter 7228, loss: 0.5517 Iter 7229, loss: 0.6010 Iter 7230, loss: 0.5657 Iter 7231, loss: 0.5770 Iter 7232, loss: 0.5510 Iter 7233, loss: 0.5972 Iter 7234, loss: 0.5687 Iter 7235, loss: 0.5353 Iter 7236, loss: 0.5520 Iter 7237, loss: 0.5512 Iter 7238, loss: 0.5732 Iter 7239, loss: 0.5452 Iter 7240, loss: 0.5243 Iter 7241, loss: 0.6101 Iter 7242, loss: 0.5681 Iter 7243, loss: 0.6069 Iter 7244, loss: 0.5525 Iter 7245, loss: 0.5270 Iter 7246, loss: 0.5394 Iter 7247, loss: 0.5331 Iter 7248, loss: 0.5367 Iter 7249, loss: 0.6075 Iter 7250, loss: 0.5641 Iter 7251, loss: 0.5367 Iter 7252, loss: 0.5544 Iter 7253, loss: 0.5073 Iter 7254, loss: 0.5384 Iter 7255, loss: 0.5623 Iter 7256, loss: 0.5464 Iter 7257, loss: 0.5566 Iter 7258, loss: 0.5459 Iter 7259, loss: 0.5692

Iter 7260, loss: 0.5540 Iter 7261, loss: 0.5691 Iter 7262, loss: 0.5743 Iter 7263, loss: 0.5290 Iter 7264, loss: 0.5295 Iter 7265, loss: 0.5870 Iter 7266, loss: 0.5664 Iter 7267, loss: 0.5509 Iter 7268, loss: 0.5631 Iter 7269, loss: 0.6175 Iter 7270, loss: 0.5660 Iter 7271, loss: 0.5156 Iter 7272, loss: 0.5606 Iter 7273, loss: 0.5952 Iter 7274, loss: 0.5504 Iter 7275, loss: 0.6100 Iter 7276, loss: 0.5131 Iter 7277, loss: 0.4893 Iter 7278, loss: 0.5465 Iter 7279, loss: 0.6044 Iter 7280, loss: 0.5710 Iter 7281, loss: 0.5682 Iter 7282, loss: 0.5421 Iter 7283, loss: 0.5349 Iter 7284, loss: 0.5755 Iter 7285, loss: 0.5325 Iter 7286, loss: 0.5679 Iter 7287, loss: 0.5754 Iter 7288, loss: 0.5838 Iter 7289, loss: 0.5707 Iter 7290, loss: 0.5800 Iter 7291, loss: 0.5808 Iter 7292, loss: 0.6095 Iter 7293, loss: 0.5378 Iter 7294, loss: 0.5483 Iter 7295, loss: 0.5635 Iter 7296, loss: 0.5846 Iter 7297, loss: 0.5405 Iter 7298, loss: 0.5555 Iter 7299, loss: 0.5776 Iter 7300, loss: 0.5654 Iter 7301, loss: 0.5483

Iter 7302, loss: 0.5470 Iter 7303, loss: 0.5253 Iter 7304, loss: 0.5864 Iter 7305, loss: 0.5837 Iter 7306, loss: 0.5172 Iter 7307, loss: 0.5884 Iter 7308, loss: 0.5531 Iter 7309, loss: 0.5189 Iter 7310, loss: 0.5556 Iter 7311, loss: 0.5470 Iter 7312, loss: 0.5308 Iter 7313, loss: 0.5317 Iter 7314, loss: 0.5609 Iter 7315, loss: 0.5849 Iter 7316, loss: 0.5735 Iter 7317, loss: 0.5631 Iter 7318, loss: 0.5364 Iter 7319, loss: 0.5633 Iter 7320, loss: 0.5468 Iter 7321, loss: 0.5567 Iter 7322, loss: 0.5151 Iter 7323, loss: 0.5257 Iter 7324, loss: 0.5227 Iter 7325, loss: 0.5967 Iter 7326, loss: 0.5382 Iter 7327, loss: 0.5177 Iter 7328, loss: 0.5599 Iter 7329, loss: 0.5326 Iter 7330, loss: 0.5149 Iter 7331, loss: 0.5354 Iter 7332, loss: 0.5460 Iter 7333, loss: 0.6162 Iter 7334, loss: 0.6046 Iter 7335, loss: 0.5320 Iter 7336, loss: 0.5739 Iter 7337, loss: 0.5475 Iter 7338, loss: 0.5811 Iter 7339, loss: 0.5874 Iter 7340, loss: 0.5833 Iter 7341, loss: 0.5403 Iter 7342, loss: 0.5620

Iter 7343, loss: 0.5402 Iter 7344, loss: 0.5570 Iter 7345, loss: 0.5280 Iter 7346, loss: 0.5600 Iter 7347, loss: 0.5644 Iter 7348, loss: 0.5576 Iter 7349, loss: 0.5497 Iter 7350, loss: 0.5608 Iter 7351, loss: 0.5807 Iter 7352, loss: 0.5471 Iter 7353, loss: 0.5778 Iter 7354, loss: 0.6017 Iter 7355, loss: 0.5938 Iter 7356, loss: 0.5589 Iter 7357, loss: 0.5173 Iter 7358, loss: 0.5648 Iter 7359, loss: 0.6122 Iter 7360, loss: 0.5627 Iter 7361, loss: 0.5113 Iter 7362, loss: 0.5205 Iter 7363, loss: 0.5629 Iter 7364, loss: 0.5529 Iter 7365, loss: 0.5583 Iter 7366, loss: 0.5862 Iter 7367, loss: 0.5331 Iter 7368, loss: 0.5813 Iter 7369, loss: 0.6057 Iter 7370, loss: 0.5419 Iter 7371, loss: 0.5641 Iter 7372, loss: 0.5911 Iter 7373, loss: 0.5736 Iter 7374, loss: 0.5766 Iter 7375, loss: 0.5261 Iter 7376, loss: 0.5533 Iter 7377, loss: 0.5920 Iter 7378, loss: 0.5704 Iter 7379, loss: 0.5525 Iter 7380, loss: 0.5914 Iter 7381, loss: 0.5541 Iter 7382, loss: 0.5103 Iter 7383, loss: 0.5360

Iter 7384, loss: 0.5370 Iter 7385, loss: 0.5072 Iter 7386, loss: 0.5537 Iter 7387, loss: 0.5464 Iter 7388, loss: 0.6019 Iter 7389, loss: 0.5534 Iter 7390, loss: 0.5666 Iter 7391, loss: 0.5485 Iter 7392, loss: 0.5466 Iter 7393, loss: 0.6001 Iter 7394, loss: 0.5641 Iter 7395, loss: 0.5786 Iter 7396, loss: 0.5190 Iter 7397, loss: 0.5588 Iter 7398, loss: 0.5297 Iter 7399, loss: 0.5776 Iter 7400, loss: 0.5737 Iter 7401, loss: 0.5799 Iter 7402, loss: 0.5444 Iter 7403, loss: 0.5886 Iter 7404, loss: 0.6426 Iter 7405, loss: 0.5657 Iter 7406, loss: 0.5712 Iter 7407, loss: 0.5544 Iter 7408, loss: 0.5609 Iter 7409, loss: 0.5630 Iter 7410, loss: 0.5577 Iter 7411, loss: 0.5310 Iter 7412, loss: 0.5117 Iter 7413, loss: 0.5510 Iter 7414, loss: 0.5500 Iter 7415, loss: 0.5483 Iter 7416, loss: 0.5774 Iter 7417, loss: 0.5410 Iter 7418, loss: 0.5609 Iter 7419, loss: 0.5146 Iter 7420, loss: 0.5605 Iter 7421, loss: 0.5195 Iter 7422, loss: 0.5536 Iter 7423, loss: 0.5532 Iter 7424, loss: 0.5298

Iter 7425, loss: 0.5619 Iter 7426, loss: 0.5890 Iter 7427, loss: 0.5785 Iter 7428, loss: 0.5483 Iter 7429, loss: 0.6357 Iter 7430, loss: 0.5411 Iter 7431, loss: 0.5448 Iter 7432, loss: 0.5702 Iter 7433, loss: 0.5622 Iter 7434, loss: 0.5666 Iter 7435, loss: 0.5406 Iter 7436, loss: 0.5610 Iter 7437, loss: 0.5183 Iter 7438, loss: 0.5376 Iter 7439, loss: 0.5547 Iter 7440, loss: 0.5741 Iter 7441, loss: 0.5769 Iter 7442, loss: 0.5704 Iter 7443, loss: 0.5542 Iter 7444, loss: 0.5377 Iter 7445, loss: 0.5551 Iter 7446, loss: 0.5627 Iter 7447, loss: 0.5445 Iter 7448, loss: 0.5564 Iter 7449, loss: 0.5485 Iter 7450, loss: 0.5274 Iter 7451, loss: 0.5577 Iter 7452, loss: 0.5630 Iter 7453, loss: 0.5788 Iter 7454, loss: 0.5003 Iter 7455, loss: 0.5557 Iter 7456, loss: 0.5451 Iter 7457, loss: 0.5501 Iter 7458, loss: 0.5234 Iter 7459, loss: 0.5094 Iter 7460, loss: 0.5136 Iter 7461, loss: 0.5253 Iter 7462, loss: 0.5442 Iter 7463, loss: 0.5232 Iter 7464, loss: 0.5523 Iter 7465, loss: 0.5886 Iter 7466, loss: 0.5818

Iter 7467, loss: 0.5543 Iter 7468, loss: 0.5648 Iter 7469, loss: 0.5411 Iter 7470, loss: 0.5746 Iter 7471, loss: 0.5939 Iter 7472, loss: 0.5440 Iter 7473, loss: 0.5649 Iter 7474, loss: 0.5662 Iter 7475, loss: 0.5626 Iter 7476, loss: 0.5524 Iter 7477, loss: 0.5586 Iter 7478, loss: 0.5695 Iter 7479, loss: 0.5423 Iter 7480, loss: 0.5135 Iter 7481, loss: 0.5293 Iter 7482, loss: 0.5311 Iter 7483, loss: 0.5931 Iter 7484, loss: 0.5761 Iter 7485, loss: 0.4821 Iter 7486, loss: 0.5169 Iter 7487, loss: 0.5652 Iter 7488, loss: 0.5523 Iter 7489, loss: 0.5618 Iter 7490, loss: 0.5118 Iter 7491, loss: 0.5795 Iter 7492, loss: 0.5344 Iter 7493, loss: 0.5442 Iter 7494, loss: 0.5436 Iter 7495, loss: 0.5149 Iter 7496, loss: 0.5393 Iter 7497, loss: 0.5258 Iter 7498, loss: 0.5090 Iter 7499, loss: 0.5775 Iter 7500, loss: 0.5272 Iter 7501, loss: 0.5013 Iter 7502, loss: 0.5762 Iter 7503, loss: 0.5258 Iter 7504, loss: 0.5196 Iter 7505, loss: 0.5836 Iter 7506, loss: 0.5329 Iter 7507, loss: 0.5858 Iter 7508, loss: 0.5223

Iter 7509, loss: 0.5826 Iter 7510, loss: 0.5556 Iter 7511, loss: 0.5674 Iter 7512, loss: 0.5367 Iter 7513, loss: 0.5594 Iter 7514, loss: 0.5792 Iter 7515, loss: 0.5203 Iter 7516, loss: 0.5705 Iter 7517, loss: 0.5087 Iter 7518, loss: 0.5498 Iter 7519, loss: 0.5331 Iter 7520, loss: 0.5371 Iter 7521, loss: 0.5389 Iter 7522, loss: 0.5427 Iter 7523, loss: 0.5734 Iter 7524, loss: 0.5930 Iter 7525, loss: 0.5370 Iter 7526, loss: 0.5237 Iter 7527, loss: 0.5969 Iter 7528, loss: 0.5891 Iter 7529, loss: 0.5253 Iter 7530, loss: 0.5732 Iter 7531, loss: 0.5519 Iter 7532, loss: 0.5529 Iter 7533, loss: 0.5776 Iter 7534, loss: 0.5595 Iter 7535, loss: 0.5160 Iter 7536, loss: 0.5166 Iter 7537, loss: 0.5266 Iter 7538, loss: 0.5817 Iter 7539, loss: 0.5189 Iter 7540, loss: 0.5504 Iter 7541, loss: 0.5156 Iter 7542, loss: 0.5581 Iter 7543, loss: 0.5431 Iter 7544, loss: 0.5698 Iter 7545, loss: 0.5607 Iter 7546, loss: 0.5702 Iter 7547, loss: 0.5800 Iter 7548, loss: 0.5757 Iter 7549, loss: 0.5247

Iter 7550, loss: 0.5611 Iter 7551, loss: 0.5253 Iter 7552, loss: 0.5067 Iter 7553, loss: 0.5623 Iter 7554, loss: 0.5252 Iter 7555, loss: 0.5599 Iter 7556, loss: 0.5863 Iter 7557, loss: 0.5164 Iter 7558, loss: 0.5748 Iter 7559, loss: 0.5258 Iter 7560, loss: 0.5639 Iter 7561, loss: 0.5451 Iter 7562, loss: 0.5676 Iter 7563, loss: 0.6020 Iter 7564, loss: 0.5521 Iter 7565, loss: 0.5227 Iter 7566, loss: 0.5680 Iter 7567, loss: 0.5354 Iter 7568, loss: 0.5448 Iter 7569, loss: 0.5544 Iter 7570, loss: 0.5737 Iter 7571, loss: 0.5958 Iter 7572, loss: 0.5602 Iter 7573, loss: 0.5697 Iter 7574, loss: 0.5727 Iter 7575, loss: 0.5285 Iter 7576, loss: 0.5364 Iter 7577, loss: 0.5601 Iter 7578, loss: 0.5255 Iter 7579, loss: 0.5393 Iter 7580, loss: 0.5591 Iter 7581, loss: 0.6079 Iter 7582, loss: 0.5293 Iter 7583, loss: 0.5344 Iter 7584, loss: 0.5134 Iter 7585, loss: 0.5747 Iter 7586, loss: 0.5396 Iter 7587, loss: 0.5783 Iter 7588, loss: 0.5716 Iter 7589, loss: 0.5878 Iter 7590, loss: 0.5349 Iter 7591, loss: 0.5643

Iter 7592, loss: 0.5531 Iter 7593, loss: 0.5064 Iter 7594, loss: 0.5774 Iter 7595, loss: 0.5688 Iter 7596, loss: 0.5237 Iter 7597, loss: 0.5645 Iter 7598, loss: 0.5919 Iter 7599, loss: 0.5685 Iter 7600, loss: 0.5625 Iter 7601, loss: 0.5845 Iter 7602, loss: 0.5420 Iter 7603, loss: 0.5349 Iter 7604, loss: 0.5252 Iter 7605, loss: 0.5145 Iter 7606, loss: 0.5735 Iter 7607, loss: 0.5924 Iter 7608, loss: 0.5155 Iter 7609, loss: 0.5439 Iter 7610, loss: 0.5837 Iter 7611, loss: 0.5553 Iter 7612, loss: 0.5484 Iter 7613, loss: 0.5377 Iter 7614, loss: 0.5599 Iter 7615, loss: 0.5593 Iter 7616, loss: 0.5641 Iter 7617, loss: 0.5601 Iter 7618, loss: 0.5386 Iter 7619, loss: 0.5314 Iter 7620, loss: 0.5348 Iter 7621, loss: 0.5480 Iter 7622, loss: 0.5805 Iter 7623, loss: 0.5444 Iter 7624, loss: 0.5396 Iter 7625, loss: 0.5404 Iter 7626, loss: 0.5752 Iter 7627, loss: 0.5582 Iter 7628, loss: 0.5592 Iter 7629, loss: 0.5423 Iter 7630, loss: 0.5371 Iter 7631, loss: 0.5751 Iter 7632, loss: 0.5433 Iter 7633, loss: 0.5748

Iter 7634, loss: 0.5673 Iter 7635, loss: 0.5693 Iter 7636, loss: 0.5602 Iter 7637, loss: 0.5488 Iter 7638, loss: 0.4945 Iter 7639, loss: 0.5727 Iter 7640, loss: 0.5104 Iter 7641, loss: 0.5292 Iter 7642, loss: 0.5470 Iter 7643, loss: 0.5672 Iter 7644, loss: 0.5475 Iter 7645, loss: 0.5960 Iter 7646, loss: 0.5024 Iter 7647, loss: 0.4973 Iter 7648, loss: 0.5298 Iter 7649, loss: 0.5792 Iter 7650, loss: 0.5336 Iter 7651, loss: 0.5563 Iter 7652, loss: 0.5816 Iter 7653, loss: 0.5737 Iter 7654, loss: 0.5673 Iter 7655, loss: 0.5617 Iter 7656, loss: 0.5769 Iter 7657, loss: 0.5598 Iter 7658, loss: 0.4959 Iter 7659, loss: 0.5392 Iter 7660, loss: 0.5257 Iter 7661, loss: 0.5556 Iter 7662, loss: 0.5450 Iter 7663, loss: 0.5230 Iter 7664, loss: 0.6097 Iter 7665, loss: 0.5345 Iter 7666, loss: 0.5396 Iter 7667, loss: 0.5456 Iter 7668, loss: 0.6041 Iter 7669, loss: 0.5526 Iter 7670, loss: 0.5793 Iter 7671, loss: 0.6184 Iter 7672, loss: 0.5807 Iter 7673, loss: 0.5305 Iter 7674, loss: 0.5653 Iter 7675, loss: 0.5253

Iter 7676, loss: 0.5695 Iter 7677, loss: 0.5100 Iter 7678, loss: 0.5386 Iter 7679, loss: 0.5471 Iter 7680, loss: 0.5611 Iter 7681, loss: 0.5376 Iter 7682, loss: 0.5362 Iter 7683, loss: 0.5253 Iter 7684, loss: 0.5699 Iter 7685, loss: 0.5535 Iter 7686, loss: 0.5607 Iter 7687, loss: 0.5195 Iter 7688, loss: 0.5724 Iter 7689, loss: 0.5538 Iter 7690, loss: 0.5397 Iter 7691, loss: 0.5761 Iter 7692, loss: 0.5536 Iter 7693, loss: 0.5865 Iter 7694, loss: 0.5427 Iter 7695, loss: 0.5478 Iter 7696, loss: 0.5549 Iter 7697, loss: 0.5592 Iter 7698, loss: 0.5066 Iter 7699, loss: 0.5771 Iter 7700, loss: 0.5315 Iter 7701, loss: 0.5550 Iter 7702, loss: 0.5427 Iter 7703, loss: 0.5463 Iter 7704, loss: 0.5445 Iter 7705, loss: 0.5346 Iter 7706, loss: 0.5616 Iter 7707, loss: 0.5481 Iter 7708, loss: 0.5759 Iter 7709, loss: 0.5413 Iter 7710, loss: 0.5235 Iter 7711, loss: 0.5369 Iter 7712, loss: 0.5304 Iter 7713, loss: 0.5343 Iter 7714, loss: 0.5524 Iter 7715, loss: 0.5384 Iter 7716, loss: 0.5076 Iter 7717, loss: 0.4992

Iter 7718, loss: 0.5993 Iter 7719, loss: 0.5243 Iter 7720, loss: 0.5837 Iter 7721, loss: 0.5548 Iter 7722, loss: 0.5677 Iter 7723, loss: 0.5493 Iter 7724, loss: 0.5388 Iter 7725, loss: 0.5794 Iter 7726, loss: 0.5314 Iter 7727, loss: 0.5589 Iter 7728, loss: 0.5556 Iter 7729, loss: 0.5648 Iter 7730, loss: 0.5303 Iter 7731, loss: 0.5282 Iter 7732, loss: 0.5375 Iter 7733, loss: 0.5409 Iter 7734, loss: 0.5405 Iter 7735, loss: 0.5336 Iter 7736, loss: 0.5589 Iter 7737, loss: 0.5107 Iter 7738, loss: 0.5877 Iter 7739, loss: 0.5329 Iter 7740, loss: 0.5253 Iter 7741, loss: 0.5253 Iter 7742, loss: 0.5619 Iter 7743, loss: 0.5461 Iter 7744, loss: 0.6001 Iter 7745, loss: 0.5935 Iter 7746, loss: 0.5593 Iter 7747, loss: 0.5409 Iter 7748, loss: 0.4985 Iter 7749, loss: 0.5438 Iter 7750, loss: 0.5420 Iter 7751, loss: 0.5711 Iter 7752, loss: 0.5600 Iter 7753, loss: 0.5483 Iter 7754, loss: 0.5335 Iter 7755, loss: 0.5624 Iter 7756, loss: 0.5256 Iter 7757, loss: 0.5572 Iter 7758, loss: 0.5427 Iter 7759, loss: 0.5372

Iter 7760, loss: 0.5657 Iter 7761, loss: 0.5328 Iter 7762, loss: 0.5318 Iter 7763, loss: 0.5701 Iter 7764, loss: 0.5400 Iter 7765, loss: 0.4964 Iter 7766, loss: 0.5336 Iter 7767, loss: 0.5509 Iter 7768, loss: 0.5364 Iter 7769, loss: 0.5796 Iter 7770, loss: 0.5493 Iter 7771, loss: 0.5367 Iter 7772, loss: 0.5293 Iter 7773, loss: 0.5536 Iter 7774, loss: 0.5317 Iter 7775, loss: 0.5100 Iter 7776, loss: 0.5804 Iter 7777, loss: 0.5048 Iter 7778, loss: 0.5246 Iter 7779, loss: 0.5638 Iter 7780, loss: 0.5037 Iter 7781, loss: 0.5317 Iter 7782, loss: 0.5646 Iter 7783, loss: 0.6096 Iter 7784, loss: 0.5879 Iter 7785, loss: 0.5466 Iter 7786, loss: 0.5227 Iter 7787, loss: 0.5550 Iter 7788, loss: 0.5342 Iter 7789, loss: 0.5599 Iter 7790, loss: 0.5025 Iter 7791, loss: 0.5598 Iter 7792, loss: 0.5618 Iter 7793, loss: 0.5736 Iter 7794, loss: 0.5511 Iter 7795, loss: 0.5536 Iter 7796, loss: 0.5524 Iter 7797, loss: 0.5226 Iter 7798, loss: 0.5730 Iter 7799, loss: 0.5026 Iter 7800, loss: 0.5833 Iter 7801, loss: 0.5761

Iter 7802, loss: 0.5995 Iter 7803, loss: 0.5665 Iter 7804, loss: 0.5562 Iter 7805, loss: 0.5460 Iter 7806, loss: 0.5122 Iter 7807, loss: 0.5729 Iter 7808, loss: 0.5562 Iter 7809, loss: 0.5248 Iter 7810, loss: 0.5838 Iter 7811, loss: 0.4985 Iter 7812, loss: 0.5692 Iter 7813, loss: 0.5045 Iter 7814, loss: 0.5233 Iter 7815, loss: 0.5526 Iter 7816, loss: 0.5428 Iter 7817, loss: 0.5485 Iter 7818, loss: 0.5632 Iter 7819, loss: 0.5260 Iter 7820, loss: 0.5539 Iter 7821, loss: 0.5445 Iter 7822, loss: 0.5250 Iter 7823, loss: 0.5532 Iter 7824, loss: 0.5230 Iter 7825, loss: 0.5494 Iter 7826, loss: 0.5586 Iter 7827, loss: 0.5086 Iter 7828, loss: 0.5555 Iter 7829, loss: 0.5450 Iter 7830, loss: 0.5581 Iter 7831, loss: 0.4870 Iter 7832, loss: 0.5573 Iter 7833, loss: 0.5893 Iter 7834, loss: 0.5178 Iter 7835, loss: 0.5572 Iter 7836, loss: 0.5108 Iter 7837, loss: 0.5551 Iter 7838, loss: 0.5481 Iter 7839, loss: 0.5437 Iter 7840, loss: 0.5498 Iter 7841, loss: 0.5277 Iter 7842, loss: 0.5622 Iter 7843, loss: 0.5637

Iter 7844, loss: 0.5838 Iter 7845, loss: 0.5464 Iter 7846, loss: 0.5215 Iter 7847, loss: 0.5158 Iter 7848, loss: 0.5193 Iter 7849, loss: 0.6126 Iter 7850, loss: 0.5815 Iter 7851, loss: 0.5251 Iter 7852, loss: 0.5097 Iter 7853, loss: 0.5745 Iter 7854, loss: 0.5599 Iter 7855, loss: 0.5195 Iter 7856, loss: 0.5845 Iter 7857, loss: 0.5482 Iter 7858, loss: 0.5787 Iter 7859, loss: 0.4780 Iter 7860, loss: 0.5914 Iter 7861, loss: 0.5866 Iter 7862, loss: 0.5345 Iter 7863, loss: 0.5771 Iter 7864, loss: 0.5489 Iter 7865, loss: 0.5729 Iter 7866, loss: 0.5308 Iter 7867, loss: 0.5305 Iter 7868, loss: 0.4968 Iter 7869, loss: 0.5167 Iter 7870, loss: 0.5541 Iter 7871, loss: 0.5744 Iter 7872, loss: 0.5770 Iter 7873, loss: 0.5564 Iter 7874, loss: 0.4900 Iter 7875, loss: 0.5439 Iter 7876, loss: 0.5328 Iter 7877, loss: 0.5699 Iter 7878, loss: 0.5654 Iter 7879, loss: 0.5330 Iter 7880, loss: 0.4999 Iter 7881, loss: 0.5298 Iter 7882, loss: 0.5396 Iter 7883, loss: 0.5527 Iter 7884, loss: 0.5432 Iter 7885, loss: 0.5268

Iter 7886, loss: 0.5839 Iter 7887, loss: 0.5295 Iter 7888, loss: 0.5771 Iter 7889, loss: 0.5787 Iter 7890, loss: 0.5173 Iter 7891, loss: 0.5094 Iter 7892, loss: 0.5201 Iter 7893, loss: 0.5705 Iter 7894, loss: 0.5350 Iter 7895, loss: 0.6088 Iter 7896, loss: 0.5752 Iter 7897, loss: 0.5105 Iter 7898, loss: 0.5454 Iter 7899, loss: 0.5530 Iter 7900, loss: 0.5481 Iter 7901, loss: 0.5757 Iter 7902, loss: 0.5697 Iter 7903, loss: 0.5212 Iter 7904, loss: 0.5340 Iter 7905, loss: 0.5441 Iter 7906, loss: 0.5183 Iter 7907, loss: 0.5066 Iter 7908, loss: 0.5538 Iter 7909, loss: 0.5890 Iter 7910, loss: 0.5036 Iter 7911, loss: 0.5625 Iter 7912, loss: 0.5367 Iter 7913, loss: 0.5688 Iter 7914, loss: 0.5498 Iter 7915, loss: 0.5564 Iter 7916, loss: 0.5544 Iter 7917, loss: 0.5412 Iter 7918, loss: 0.4839 Iter 7919, loss: 0.5176 Iter 7920, loss: 0.5328 Iter 7921, loss: 0.5389 Iter 7922, loss: 0.5399 Iter 7923, loss: 0.5364 Iter 7924, loss: 0.5495 Iter 7925, loss: 0.5469 Iter 7926, loss: 0.5468 Iter 7927, loss: 0.5261

Iter 7928, loss: 0.5481 Iter 7929, loss: 0.5326 Iter 7930, loss: 0.5374 Iter 7931, loss: 0.5607 Iter 7932, loss: 0.5566 Iter 7933, loss: 0.5621 Iter 7934, loss: 0.5659 Iter 7935, loss: 0.5399 Iter 7936, loss: 0.5590 Iter 7937, loss: 0.5740 Iter 7938, loss: 0.4657 Iter 7939, loss: 0.5363 Iter 7940, loss: 0.5690 Iter 7941, loss: 0.5677 Iter 7942, loss: 0.5067 Iter 7943, loss: 0.5492 Iter 7944, loss: 0.5483 Iter 7945, loss: 0.5427 Iter 7946, loss: 0.5291 Iter 7947, loss: 0.5141 Iter 7948, loss: 0.5542 Iter 7949, loss: 0.5905 Iter 7950, loss: 0.5769 Iter 7951, loss: 0.5384 Iter 7952, loss: 0.5718 Iter 7953, loss: 0.5287 Iter 7954, loss: 0.5199 Iter 7955, loss: 0.5132 Iter 7956, loss: 0.5333 Iter 7957, loss: 0.5113 Iter 7958, loss: 0.5583 Iter 7959, loss: 0.5279 Iter 7960, loss: 0.5575 Iter 7961, loss: 0.5106 Iter 7962, loss: 0.5257 Iter 7963, loss: 0.5192 Iter 7964, loss: 0.5414 Iter 7965, loss: 0.5023 Iter 7966, loss: 0.4972 Iter 7967, loss: 0.5408 Iter 7968, loss: 0.5191 Iter 7969, loss: 0.5495

Iter 7970, loss: 0.5195 Iter 7971, loss: 0.5524 Iter 7972, loss: 0.5452 Iter 7973, loss: 0.5562 Iter 7974, loss: 0.5968 Iter 7975, loss: 0.5447 Iter 7976, loss: 0.5708 Iter 7977, loss: 0.5561 Iter 7978, loss: 0.5297 Iter 7979, loss: 0.5790 Iter 7980, loss: 0.5453 Iter 7981, loss: 0.5594 Iter 7982, loss: 0.5402 Iter 7983, loss: 0.5199 Iter 7984, loss: 0.5830 Iter 7985, loss: 0.5289 Iter 7986, loss: 0.5643 Iter 7987, loss: 0.5707 Iter 7988, loss: 0.5402 Iter 7989, loss: 0.5133 Iter 7990, loss: 0.5462 Iter 7991, loss: 0.5042 Iter 7992, loss: 0.5361 Iter 7993, loss: 0.5553 Iter 7994, loss: 0.5772 Iter 7995, loss: 0.5330 Iter 7996, loss: 0.5568 Iter 7997, loss: 0.5353 Iter 7998, loss: 0.5038 Iter 7999, loss: 0.5957 Iter 8000, loss: 0.5421 Iter 8001, loss: 0.5709 Iter 8002, loss: 0.5869 Iter 8003, loss: 0.5426 Iter 8004, loss: 0.5570 Iter 8005, loss: 0.5445 Iter 8006, loss: 0.5311 Iter 8007, loss: 0.4815 Iter 8008, loss: 0.4883 Iter 8009, loss: 0.5617 Iter 8010, loss: 0.5277 Iter 8011, loss: 0.5285

Iter 8012, loss: 0.5851 Iter 8013, loss: 0.5140 Iter 8014, loss: 0.5313 Iter 8015, loss: 0.5818 Iter 8016, loss: 0.5264 Iter 8017, loss: 0.5298 Iter 8018, loss: 0.5790 Iter 8019, loss: 0.5809 Iter 8020, loss: 0.5533 Iter 8021, loss: 0.5250 Iter 8022, loss: 0.5133 Iter 8023, loss: 0.5570 Iter 8024, loss: 0.5608 Iter 8025, loss: 0.5850 Iter 8026, loss: 0.5849 Iter 8027, loss: 0.5798 Iter 8028, loss: 0.5726 Iter 8029, loss: 0.5227 Iter 8030, loss: 0.5022 Iter 8031, loss: 0.5425 Iter 8032, loss: 0.5710 Iter 8033, loss: 0.5404 Iter 8034, loss: 0.5232 Iter 8035, loss: 0.5706 Iter 8036, loss: 0.5730 Iter 8037, loss: 0.5336 Iter 8038, loss: 0.5195 Iter 8039, loss: 0.5806 Iter 8040, loss: 0.5060 Iter 8041, loss: 0.5252 Iter 8042, loss: 0.5457 Iter 8043, loss: 0.5385 Iter 8044, loss: 0.5471 Iter 8045, loss: 0.5125 Iter 8046, loss: 0.5468 Iter 8047, loss: 0.5285 Iter 8048, loss: 0.5501 Iter 8049, loss: 0.5346 Iter 8050, loss: 0.5508 Iter 8051, loss: 0.5178 Iter 8052, loss: 0.5375

Iter 8053, loss: 0.4700 Iter 8054, loss: 0.5586 Iter 8055, loss: 0.5522 Iter 8056, loss: 0.5456 Iter 8057, loss: 0.5586 Iter 8058, loss: 0.5425 Iter 8059, loss: 0.5302 Iter 8060, loss: 0.5040 Iter 8061, loss: 0.5630 Iter 8062, loss: 0.5683 Iter 8063, loss: 0.5416 Iter 8064, loss: 0.5302 Iter 8065, loss: 0.5415 Iter 8066, loss: 0.5472 Iter 8067, loss: 0.5256 Iter 8068, loss: 0.5282 Iter 8069, loss: 0.5294 Iter 8070, loss: 0.5432 Iter 8071, loss: 0.5166 Iter 8072, loss: 0.5487 Iter 8073, loss: 0.5270 Iter 8074, loss: 0.5788 Iter 8075, loss: 0.4727 Iter 8076, loss: 0.5548 Iter 8077, loss: 0.5403 Iter 8078, loss: 0.5126 Iter 8079, loss: 0.4998 Iter 8080, loss: 0.5594 Iter 8081, loss: 0.5671 Iter 8082, loss: 0.5685 Iter 8083, loss: 0.5086 Iter 8084, loss: 0.5478 Iter 8085, loss: 0.5398 Iter 8086, loss: 0.5499 Iter 8087, loss: 0.5374 Iter 8088, loss: 0.5106 Iter 8089, loss: 0.5481 Iter 8090, loss: 0.5289 Iter 8091, loss: 0.5218 Iter 8092, loss: 0.5665 Iter 8093, loss: 0.6042 Iter 8094, loss: 0.5445

Iter 8095, loss: 0.5540 Iter 8096, loss: 0.5733 Iter 8097, loss: 0.5418 Iter 8098, loss: 0.5557 Iter 8099, loss: 0.5443 Iter 8100, loss: 0.6158 Iter 8101, loss: 0.5338 Iter 8102, loss: 0.5396 Iter 8103, loss: 0.5378 Iter 8104, loss: 0.5563 Iter 8105, loss: 0.5848 Iter 8106, loss: 0.5559 Iter 8107, loss: 0.5448 Iter 8108, loss: 0.5733 Iter 8109, loss: 0.5520 Iter 8110, loss: 0.5623 Iter 8111, loss: 0.5131 Iter 8112, loss: 0.5416 Iter 8113, loss: 0.5423 Iter 8114, loss: 0.5861 Iter 8115, loss: 0.5528 Iter 8116, loss: 0.5430 Iter 8117, loss: 0.5658 Iter 8118, loss: 0.5031 Iter 8119, loss: 0.5392 Iter 8120, loss: 0.5175 Iter 8121, loss: 0.5583 Iter 8122, loss: 0.5690 Iter 8123, loss: 0.5229 Iter 8124, loss: 0.5204 Iter 8125, loss: 0.5364 Iter 8126, loss: 0.5293 Iter 8127, loss: 0.5239 Iter 8128, loss: 0.5629 Iter 8129, loss: 0.5600 Iter 8130, loss: 0.5109 Iter 8131, loss: 0.5577 Iter 8132, loss: 0.5054 Iter 8133, loss: 0.5579 Iter 8134, loss: 0.5222 Iter 8135, loss: 0.5435 Iter 8136, loss: 0.5463

Iter 8137, loss: 0.5410 Iter 8138, loss: 0.5516 Iter 8139, loss: 0.5447 Iter 8140, loss: 0.6064 Iter 8141, loss: 0.5552 Iter 8142, loss: 0.5756 Iter 8143, loss: 0.5258 Iter 8144, loss: 0.5316 Iter 8145, loss: 0.5528 Iter 8146, loss: 0.5232 Iter 8147, loss: 0.5636 Iter 8148, loss: 0.4902 Iter 8149, loss: 0.5362 Iter 8150, loss: 0.5275 Iter 8151, loss: 0.5323 Iter 8152, loss: 0.5553 Iter 8153, loss: 0.5410 Iter 8154, loss: 0.5593 Iter 8155, loss: 0.5508 Iter 8156, loss: 0.5544 Iter 8157, loss: 0.4988 Iter 8158, loss: 0.5701 Iter 8159, loss: 0.5330 Iter 8160, loss: 0.5227 Iter 8161, loss: 0.5374 Iter 8162, loss: 0.5379 Iter 8163, loss: 0.5477 Iter 8164, loss: 0.5416 Iter 8165, loss: 0.5862 Iter 8166, loss: 0.5464 Iter 8167, loss: 0.5224 Iter 8168, loss: 0.5371 Iter 8169, loss: 0.5330 Iter 8170, loss: 0.5746 Iter 8171, loss: 0.5466 Iter 8172, loss: 0.5248 Iter 8173, loss: 0.5100 Iter 8174, loss: 0.5777 Iter 8175, loss: 0.4865 Iter 8176, loss: 0.5537 Iter 8177, loss: 0.5059 Iter 8178, loss: 0.5876

Iter 8179, loss: 0.5229 Iter 8180, loss: 0.5465 Iter 8181, loss: 0.5214 Iter 8182, loss: 0.5297 Iter 8183, loss: 0.5470 Iter 8184, loss: 0.5478 Iter 8185, loss: 0.5307 Iter 8186, loss: 0.5548 Iter 8187, loss: 0.5424 Iter 8188, loss: 0.5691 Iter 8189, loss: 0.5643 Iter 8190, loss: 0.5054 Iter 8191, loss: 0.5123 Iter 8192, loss: 0.5834 Iter 8193, loss: 0.5114 Iter 8194, loss: 0.5217 Iter 8195, loss: 0.5076 Iter 8196, loss: 0.5406 Iter 8197, loss: 0.5861 Iter 8198, loss: 0.5040 Iter 8199, loss: 0.5496 Iter 8200, loss: 0.5287 Iter 8201, loss: 0.4860 Iter 8202, loss: 0.5568 Iter 8203, loss: 0.5153 Iter 8204, loss: 0.5683 Iter 8205, loss: 0.5361 Iter 8206, loss: 0.5836 Iter 8207, loss: 0.5437 Iter 8208, loss: 0.5548 Iter 8209, loss: 0.4928 Iter 8210, loss: 0.5636 Iter 8211, loss: 0.5053 Iter 8212, loss: 0.5421 Iter 8213, loss: 0.5382 Iter 8214, loss: 0.5641 Iter 8215, loss: 0.5079 Iter 8216, loss: 0.5493 Iter 8217, loss: 0.5540 Iter 8218, loss: 0.5117 Iter 8219, loss: 0.5232

Iter 8220, loss: 0.5332 Iter 8221, loss: 0.5223 Iter 8222, loss: 0.5287 Iter 8223, loss: 0.5562 Iter 8224, loss: 0.5204 Iter 8225, loss: 0.5149 Iter 8226, loss: 0.5563 Iter 8227, loss: 0.5570 Iter 8228, loss: 0.5081 Iter 8229, loss: 0.5804 Iter 8230, loss: 0.5510 Iter 8231, loss: 0.5814 Iter 8232, loss: 0.5608 Iter 8233, loss: 0.5026 Iter 8234, loss: 0.5574 Iter 8235, loss: 0.5532 Iter 8236, loss: 0.5563 Iter 8237, loss: 0.5457 Iter 8238, loss: 0.5997 Iter 8239, loss: 0.5485 Iter 8240, loss: 0.5457 Iter 8241, loss: 0.5387 Iter 8242, loss: 0.5404 Iter 8243, loss: 0.5485 Iter 8244, loss: 0.5230 Iter 8245, loss: 0.5349 Iter 8246, loss: 0.5302 Iter 8247, loss: 0.5226 Iter 8248, loss: 0.5869 Iter 8249, loss: 0.5818 Iter 8250, loss: 0.5674 Iter 8251, loss: 0.5495 Iter 8252, loss: 0.5693 Iter 8253, loss: 0.5640 Iter 8254, loss: 0.5198 Iter 8255, loss: 0.5461 Iter 8256, loss: 0.5463 Iter 8257, loss: 0.5587 Iter 8258, loss: 0.5955 Iter 8259, loss: 0.5314 Iter 8260, loss: 0.5416 Iter 8261, loss: 0.5764

Iter 8262, loss: 0.5471 Iter 8263, loss: 0.5490 Iter 8264, loss: 0.5531 Iter 8265, loss: 0.5528 Iter 8266, loss: 0.5524 Iter 8267, loss: 0.5614 Iter 8268, loss: 0.5714 Iter 8269, loss: 0.5186 Iter 8270, loss: 0.5618 Iter 8271, loss: 0.5489 Iter 8272, loss: 0.5083 Iter 8273, loss: 0.5340 Iter 8274, loss: 0.5357 Iter 8275, loss: 0.5401 Iter 8276, loss: 0.5685 Iter 8277, loss: 0.5231 Iter 8278, loss: 0.5778 Iter 8279, loss: 0.5489 Iter 8280, loss: 0.5319 Iter 8281, loss: 0.5076 Iter 8282, loss: 0.5342 Iter 8283, loss: 0.5179 Iter 8284, loss: 0.5207 Iter 8285, loss: 0.5748 Iter 8286, loss: 0.5914 Iter 8287, loss: 0.5650 Iter 8288, loss: 0.5550 Iter 8289, loss: 0.5533 Iter 8290, loss: 0.5570 Iter 8291, loss: 0.5422 Iter 8292, loss: 0.5615 Iter 8293, loss: 0.5641 Iter 8294, loss: 0.5372 Iter 8295, loss: 0.5263 Iter 8296, loss: 0.5539 Iter 8297, loss: 0.5745 Iter 8298, loss: 0.5350 Iter 8299, loss: 0.5212 Iter 8300, loss: 0.4911 Iter 8301, loss: 0.5361 Iter 8302, loss: 0.5186

Iter 8303, loss: 0.5067 Iter 8304, loss: 0.5193 Iter 8305, loss: 0.5463 Iter 8306, loss: 0.5390 Iter 8307, loss: 0.5225 Iter 8308, loss: 0.5313 Iter 8309, loss: 0.5505 Iter 8310, loss: 0.5592 Iter 8311, loss: 0.5603 Iter 8312, loss: 0.5471 Iter 8313, loss: 0.5664 Iter 8314, loss: 0.5062 Iter 8315, loss: 0.5526 Iter 8316, loss: 0.5821 Iter 8317, loss: 0.5172 Iter 8318, loss: 0.5665 Iter 8319, loss: 0.5592 Iter 8320, loss: 0.5333 Iter 8321, loss: 0.5250 Iter 8322, loss: 0.5528 Iter 8323, loss: 0.5623 Iter 8324, loss: 0.5173 Iter 8325, loss: 0.5516 Iter 8326, loss: 0.5518 Iter 8327, loss: 0.5595 Iter 8328, loss: 0.4868 Iter 8329, loss: 0.5317 Iter 8330, loss: 0.5231 Iter 8331, loss: 0.5193 Iter 8332, loss: 0.5234 Iter 8333, loss: 0.5559 Iter 8334, loss: 0.5273 Iter 8335, loss: 0.5436 Iter 8336, loss: 0.5124 Iter 8337, loss: 0.5472 Iter 8338, loss: 0.5353 Iter 8339, loss: 0.5729 Iter 8340, loss: 0.5838 Iter 8341, loss: 0.5136 Iter 8342, loss: 0.5393 Iter 8343, loss: 0.5254

Iter 8344, loss: 0.5380 Iter 8345, loss: 0.5812 Iter 8346, loss: 0.5330 Iter 8347, loss: 0.5369 Iter 8348, loss: 0.5283 Iter 8349, loss: 0.5572 Iter 8350, loss: 0.5464 Iter 8351, loss: 0.5356 Iter 8352, loss: 0.5373 Iter 8353, loss: 0.5831 Iter 8354, loss: 0.5529 Iter 8355, loss: 0.5391 Iter 8356, loss: 0.5685 Iter 8357, loss: 0.5303 Iter 8358, loss: 0.5410 Iter 8359, loss: 0.5079 Iter 8360, loss: 0.5523 Iter 8361, loss: 0.5503 Iter 8362, loss: 0.5588 Iter 8363, loss: 0.5369 Iter 8364, loss: 0.5400 Iter 8365, loss: 0.5631 Iter 8366, loss: 0.5339 Iter 8367, loss: 0.5597 Iter 8368, loss: 0.5345 Iter 8369, loss: 0.5268 Iter 8370, loss: 0.5379 Iter 8371, loss: 0.5839 Iter 8372, loss: 0.5243 Iter 8373, loss: 0.5328 Iter 8374, loss: 0.5344 Iter 8375, loss: 0.4702 Iter 8376, loss: 0.5834 Iter 8377, loss: 0.5757 Iter 8378, loss: 0.4974 Iter 8379, loss: 0.5172 Iter 8380, loss: 0.5413 Iter 8381, loss: 0.5618 Iter 8382, loss: 0.5513 Iter 8383, loss: 0.5570 Iter 8384, loss: 0.4910 Iter 8385, loss: 0.5295

Iter 8386, loss: 0.5586 Iter 8387, loss: 0.5532 Iter 8388, loss: 0.5111 Iter 8389, loss: 0.5849 Iter 8390, loss: 0.5372 Iter 8391, loss: 0.5418 Iter 8392, loss: 0.5676 Iter 8393, loss: 0.5167 Iter 8394, loss: 0.5157 Iter 8395, loss: 0.5253 Iter 8396, loss: 0.5467 Iter 8397, loss: 0.5043 Iter 8398, loss: 0.5513 Iter 8399, loss: 0.5188 Iter 8400, loss: 0.5271 Iter 8401, loss: 0.5884 Iter 8402, loss: 0.5572 Iter 8403, loss: 0.6075 Iter 8404, loss: 0.5102 Iter 8405, loss: 0.5116 Iter 8406, loss: 0.5328 Iter 8407, loss: 0.5648 Iter 8408, loss: 0.5431 Iter 8409, loss: 0.5648 Iter 8410, loss: 0.4895 Iter 8411, loss: 0.5716 Iter 8412, loss: 0.5131 Iter 8413, loss: 0.5033 Iter 8414, loss: 0.5278 Iter 8415, loss: 0.5138 Iter 8416, loss: 0.5529 Iter 8417, loss: 0.5529 Iter 8418, loss: 0.5358 Iter 8419, loss: 0.5709 Iter 8420, loss: 0.5271 Iter 8421, loss: 0.5459 Iter 8422, loss: 0.5668 Iter 8423, loss: 0.5263 Iter 8424, loss: 0.5755 Iter 8425, loss: 0.5046 Iter 8426, loss: 0.5136

Iter 8427, loss: 0.5905 Iter 8428, loss: 0.5422 Iter 8429, loss: 0.5426 Iter 8430, loss: 0.5431 Iter 8431, loss: 0.5168 Iter 8432, loss: 0.5560 Iter 8433, loss: 0.5163 Iter 8434, loss: 0.5635 Iter 8435, loss: 0.5259 Iter 8436, loss: 0.5583 Iter 8437, loss: 0.5588 Iter 8438, loss: 0.5526 Iter 8439, loss: 0.5629 Iter 8440, loss: 0.5159 Iter 8441, loss: 0.5933 Iter 8442, loss: 0.5463 Iter 8443, loss: 0.5330 Iter 8444, loss: 0.5373 Iter 8445, loss: 0.5756 Iter 8446, loss: 0.5366 Iter 8447, loss: 0.5744 Iter 8448, loss: 0.5422 Iter 8449, loss: 0.5297 Iter 8450, loss: 0.5450 Iter 8451, loss: 0.4572 Iter 8452, loss: 0.5523 Iter 8453, loss: 0.5480 Iter 8454, loss: 0.5524 Iter 8455, loss: 0.5400 Iter 8456, loss: 0.5560 Iter 8457, loss: 0.5193 Iter 8458, loss: 0.5449 Iter 8459, loss: 0.5373 Iter 8460, loss: 0.5210 Iter 8461, loss: 0.4926 Iter 8462, loss: 0.5227 Iter 8463, loss: 0.5193 Iter 8464, loss: 0.5267 Iter 8465, loss: 0.5597 Iter 8466, loss: 0.5196 Iter 8467, loss: 0.5168 Iter 8468, loss: 0.5481

Iter 8469, loss: 0.4870 Iter 8470, loss: 0.5064 Iter 8471, loss: 0.5665 Iter 8472, loss: 0.5828 Iter 8473, loss: 0.5256 Iter 8474, loss: 0.5204 Iter 8475, loss: 0.5418 Iter 8476, loss: 0.5636 Iter 8477, loss: 0.5323 Iter 8478, loss: 0.5351 Iter 8479, loss: 0.5127 Iter 8480, loss: 0.5180 Iter 8481, loss: 0.5366 Iter 8482, loss: 0.5304 Iter 8483, loss: 0.5203 Iter 8484, loss: 0.5494 Iter 8485, loss: 0.5145 Iter 8486, loss: 0.5431 Iter 8487, loss: 0.5109 Iter 8488, loss: 0.5224 Iter 8489, loss: 0.5638 Iter 8490, loss: 0.5210 Iter 8491, loss: 0.5204 Iter 8492, loss: 0.5668 Iter 8493, loss: 0.5382 Iter 8494, loss: 0.5568 Iter 8495, loss: 0.6175 Iter 8496, loss: 0.5494 Iter 8497, loss: 0.5476 Iter 8498, loss: 0.5332 Iter 8499, loss: 0.5067 Iter 8500, loss: 0.5155 Iter 8501, loss: 0.5257 Iter 8502, loss: 0.5376 Iter 8503, loss: 0.5142 Iter 8504, loss: 0.5307 Iter 8505, loss: 0.5499 Iter 8506, loss: 0.5527 Iter 8507, loss: 0.5004 Iter 8508, loss: 0.5474 Iter 8509, loss: 0.5493 Iter 8510, loss: 0.5647

Iter 8511, loss: 0.5596 Iter 8512, loss: 0.5740 Iter 8513, loss: 0.5521 Iter 8514, loss: 0.5222 Iter 8515, loss: 0.5822 Iter 8516, loss: 0.5022 Iter 8517, loss: 0.5190 Iter 8518, loss: 0.5477 Iter 8519, loss: 0.5370 Iter 8520, loss: 0.5392 Iter 8521, loss: 0.5280 Iter 8522, loss: 0.5284 Iter 8523, loss: 0.5160 Iter 8524, loss: 0.5678 Iter 8525, loss: 0.4948 Iter 8526, loss: 0.5519 Iter 8527, loss: 0.5065 Iter 8528, loss: 0.4982 Iter 8529, loss: 0.5739 Iter 8530, loss: 0.5259 Iter 8531, loss: 0.5099 Iter 8532, loss: 0.4998 Iter 8533, loss: 0.5511 Iter 8534, loss: 0.5596 Iter 8535, loss: 0.5019 Iter 8536, loss: 0.5201 Iter 8537, loss: 0.5562 Iter 8538, loss: 0.5408 Iter 8539, loss: 0.5451 Iter 8540, loss: 0.5548 Iter 8541, loss: 0.5372 Iter 8542, loss: 0.5464 Iter 8543, loss: 0.4810 Iter 8544, loss: 0.5133 Iter 8545, loss: 0.5053 Iter 8546, loss: 0.5325 Iter 8547, loss: 0.5454 Iter 8548, loss: 0.5422 Iter 8549, loss: 0.5514 Iter 8550, loss: 0.5441 Iter 8551, loss: 0.5463

Iter 8552, loss: 0.5340 Iter 8553, loss: 0.5126 Iter 8554, loss: 0.5525 Iter 8555, loss: 0.5305 Iter 8556, loss: 0.5738 Iter 8557, loss: 0.5326 Iter 8558, loss: 0.5326 Iter 8559, loss: 0.5654 Iter 8560, loss: 0.5807 Iter 8561, loss: 0.5643 Iter 8562, loss: 0.4911 Iter 8563, loss: 0.5269 Iter 8564, loss: 0.5152 Iter 8565, loss: 0.5284 Iter 8566, loss: 0.5713 Iter 8567, loss: 0.5233 Iter 8568, loss: 0.5086 Iter 8569, loss: 0.5102 Iter 8570, loss: 0.5469 Iter 8571, loss: 0.5535 Iter 8572, loss: 0.5120 Iter 8573, loss: 0.5721 Iter 8574, loss: 0.5181 Iter 8575, loss: 0.5283 Iter 8576, loss: 0.5134 Iter 8577, loss: 0.5409 Iter 8578, loss: 0.5303 Iter 8579, loss: 0.5407 Iter 8580, loss: 0.5392 Iter 8581, loss: 0.5260 Iter 8582, loss: 0.5149 Iter 8583, loss: 0.5934 Iter 8584, loss: 0.5752 Iter 8585, loss: 0.5224 Iter 8586, loss: 0.5317 Iter 8587, loss: 0.5192 Iter 8588, loss: 0.5340 Iter 8589, loss: 0.5334 Iter 8590, loss: 0.5728 Iter 8591, loss: 0.5702 Iter 8592, loss: 0.5191 Iter 8593, loss: 0.5476

Iter 8594, loss: 0.5333 Iter 8595, loss: 0.5033 Iter 8596, loss: 0.5658 Iter 8597, loss: 0.5734 Iter 8598, loss: 0.5158 Iter 8599, loss: 0.5470 Iter 8600, loss: 0.5449 Iter 8601, loss: 0.5508 Iter 8602, loss: 0.5615 Iter 8603, loss: 0.5697 Iter 8604, loss: 0.5107 Iter 8605, loss: 0.5528 Iter 8606, loss: 0.5386 Iter 8607, loss: 0.5765 Iter 8608, loss: 0.5464 Iter 8609, loss: 0.5178 Iter 8610, loss: 0.5600 Iter 8611, loss: 0.5726 Iter 8612, loss: 0.5614 Iter 8613, loss: 0.5431 Iter 8614, loss: 0.5857 Iter 8615, loss: 0.5582 Iter 8616, loss: 0.5573 Iter 8617, loss: 0.5719 Iter 8618, loss: 0.5428 Iter 8619, loss: 0.5265 Iter 8620, loss: 0.5507 Iter 8621, loss: 0.5619 Iter 8622, loss: 0.5361 Iter 8623, loss: 0.5025 Iter 8624, loss: 0.5262 Iter 8625, loss: 0.4985 Iter 8626, loss: 0.5349 Iter 8627, loss: 0.5755 Iter 8628, loss: 0.5722 Iter 8629, loss: 0.5502 Iter 8630, loss: 0.5541 Iter 8631, loss: 0.5328 Iter 8632, loss: 0.5165 Iter 8633, loss: 0.5470 Iter 8634, loss: 0.5280

Iter 8635, loss: 0.5534 Iter 8636, loss: 0.5131 Iter 8637, loss: 0.5646 Iter 8638, loss: 0.5312 Iter 8639, loss: 0.5560 Iter 8640, loss: 0.5322 Iter 8641, loss: 0.5204 Iter 8642, loss: 0.5323 Iter 8643, loss: 0.5206 Iter 8644, loss: 0.5285 Iter 8645, loss: 0.5189 Iter 8646, loss: 0.5426 Iter 8647, loss: 0.5516 Iter 8648, loss: 0.5538 Iter 8649, loss: 0.4903 Iter 8650, loss: 0.4916 Iter 8651, loss: 0.4973 Iter 8652, loss: 0.5325 Iter 8653, loss: 0.5860 Iter 8654, loss: 0.5226 Iter 8655, loss: 0.5206 Iter 8656, loss: 0.5297 Iter 8657, loss: 0.5427 Iter 8658, loss: 0.5637 Iter 8659, loss: 0.5015 Iter 8660, loss: 0.5209 Iter 8661, loss: 0.5922 Iter 8662, loss: 0.5204 Iter 8663, loss: 0.5222 Iter 8664, loss: 0.5455 Iter 8665, loss: 0.5323 Iter 8666, loss: 0.4918 Iter 8667, loss: 0.5330 Iter 8668, loss: 0.5548 Iter 8669, loss: 0.5376 Iter 8670, loss: 0.5087 Iter 8671, loss: 0.4981 Iter 8672, loss: 0.5402 Iter 8673, loss: 0.5620 Iter 8674, loss: 0.5376 Iter 8675, loss: 0.5133 Iter 8676, loss: 0.5415

Iter 8677, loss: 0.5592 Iter 8678, loss: 0.5216 Iter 8679, loss: 0.5879 Iter 8680, loss: 0.5684 Iter 8681, loss: 0.5322 Iter 8682, loss: 0.5325 Iter 8683, loss: 0.5283 Iter 8684, loss: 0.5692 Iter 8685, loss: 0.5874 Iter 8686, loss: 0.5404 Iter 8687, loss: 0.5031 Iter 8688, loss: 0.5258 Iter 8689, loss: 0.5536 Iter 8690, loss: 0.5786 Iter 8691, loss: 0.5626 Iter 8692, loss: 0.5066 Iter 8693, loss: 0.5436 Iter 8694, loss: 0.5329 Iter 8695, loss: 0.5378 Iter 8696, loss: 0.5341 Iter 8697, loss: 0.5192 Iter 8698, loss: 0.5142 Iter 8699, loss: 0.5170 Iter 8700, loss: 0.5565 Iter 8701, loss: 0.5657 Iter 8702, loss: 0.5767 Iter 8703, loss: 0.5830 Iter 8704, loss: 0.5149 Iter 8705, loss: 0.5658 Iter 8706, loss: 0.5246 Iter 8707, loss: 0.5038 Iter 8708, loss: 0.4659 Iter 8709, loss: 0.5366 Iter 8710, loss: 0.5398 Iter 8711, loss: 0.5180 Iter 8712, loss: 0.5719 Iter 8713, loss: 0.5072 Iter 8714, loss: 0.4993 Iter 8715, loss: 0.5272 Iter 8716, loss: 0.5348 Iter 8717, loss: 0.5816 Iter 8718, loss: 0.4950

Iter 8719, loss: 0.4970 Iter 8720, loss: 0.6154 Iter 8721, loss: 0.4977 Iter 8722, loss: 0.5334 Iter 8723, loss: 0.4977 Iter 8724, loss: 0.5355 Iter 8725, loss: 0.5537 Iter 8726, loss: 0.5679 Iter 8727, loss: 0.5363 Iter 8728, loss: 0.5234 Iter 8729, loss: 0.5474 Iter 8730, loss: 0.4878 Iter 8731, loss: 0.5273 Iter 8732, loss: 0.4998 Iter 8733, loss: 0.5397 Iter 8734, loss: 0.5556 Iter 8735, loss: 0.5319 Iter 8736, loss: 0.5531 Iter 8737, loss: 0.5566 Iter 8738, loss: 0.5334 Iter 8739, loss: 0.5685 Iter 8740, loss: 0.5293 Iter 8741, loss: 0.5210 Iter 8742, loss: 0.5382 Iter 8743, loss: 0.5509 Iter 8744, loss: 0.5432 Iter 8745, loss: 0.4697 Iter 8746, loss: 0.5504 Iter 8747, loss: 0.5324 Iter 8748, loss: 0.5055 Iter 8749, loss: 0.5228 Iter 8750, loss: 0.5098 Iter 8751, loss: 0.5171 Iter 8752, loss: 0.5324 Iter 8753, loss: 0.5091 Iter 8754, loss: 0.5144 Iter 8755, loss: 0.5181 Iter 8756, loss: 0.5695 Iter 8757, loss: 0.5679 Iter 8758, loss: 0.5516 Iter 8759, loss: 0.5434

Iter 8760, loss: 0.5138 Iter 8761, loss: 0.4660 Iter 8762, loss: 0.5078 Iter 8763, loss: 0.5767 Iter 8764, loss: 0.5528 Iter 8765, loss: 0.5299 Iter 8766, loss: 0.5152 Iter 8767, loss: 0.5519 Iter 8768, loss: 0.5218 Iter 8769, loss: 0.5104 Iter 8770, loss: 0.5372 Iter 8771, loss: 0.5264 Iter 8772, loss: 0.5395 Iter 8773, loss: 0.5442 Iter 8774, loss: 0.5169 Iter 8775, loss: 0.5373 Iter 8776, loss: 0.5395 Iter 8777, loss: 0.5458 Iter 8778, loss: 0.5169 Iter 8779, loss: 0.5455 Iter 8780, loss: 0.5257 Iter 8781, loss: 0.5419 Iter 8782, loss: 0.5440 Iter 8783, loss: 0.5053 Iter 8784, loss: 0.5017 Iter 8785, loss: 0.5390 Iter 8786, loss: 0.5455 Iter 8787, loss: 0.5155 Iter 8788, loss: 0.5449 Iter 8789, loss: 0.5305 Iter 8790, loss: 0.5027 Iter 8791, loss: 0.5350 Iter 8792, loss: 0.5601 Iter 8793, loss: 0.4798 Iter 8794, loss: 0.5291 Iter 8795, loss: 0.5125 Iter 8796, loss: 0.5209 Iter 8797, loss: 0.4835 Iter 8798, loss: 0.5058 Iter 8799, loss: 0.5601 Iter 8800, loss: 0.5596 Iter 8801, loss: 0.5706

Iter 8802, loss: 0.5667 Iter 8803, loss: 0.5133 Iter 8804, loss: 0.5310 Iter 8805, loss: 0.5269 Iter 8806, loss: 0.5047 Iter 8807, loss: 0.5005 Iter 8808, loss: 0.5333 Iter 8809, loss: 0.5534 Iter 8810, loss: 0.5742 Iter 8811, loss: 0.5415 Iter 8812, loss: 0.5505 Iter 8813, loss: 0.5689 Iter 8814, loss: 0.4813 Iter 8815, loss: 0.5166 Iter 8816, loss: 0.4977 Iter 8817, loss: 0.5376 Iter 8818, loss: 0.5463 Iter 8819, loss: 0.4976 Iter 8820, loss: 0.5088 Iter 8821, loss: 0.5377 Iter 8822, loss: 0.5082 Iter 8823, loss: 0.5617 Iter 8824, loss: 0.5306 Iter 8825, loss: 0.5207 Iter 8826, loss: 0.5140 Iter 8827, loss: 0.5231 Iter 8828, loss: 0.5185 Iter 8829, loss: 0.5520 Iter 8830, loss: 0.5354 Iter 8831, loss: 0.5250 Iter 8832, loss: 0.5437 Iter 8833, loss: 0.5568 Iter 8834, loss: 0.5406 Iter 8835, loss: 0.5199 Iter 8836, loss: 0.5668 Iter 8837, loss: 0.5448 Iter 8838, loss: 0.5229 Iter 8839, loss: 0.5366 Iter 8840, loss: 0.5291 Iter 8841, loss: 0.5324 Iter 8842, loss: 0.5804 Iter 8843, loss: 0.4929

Iter 8844, loss: 0.5438 Iter 8845, loss: 0.5194 Iter 8846, loss: 0.5294 Iter 8847, loss: 0.5328 Iter 8848, loss: 0.5120 Iter 8849, loss: 0.5199 Iter 8850, loss: 0.5215 Iter 8851, loss: 0.5259 Iter 8852, loss: 0.5198 Iter 8853, loss: 0.5211 Iter 8854, loss: 0.5592 Iter 8855, loss: 0.4896 Iter 8856, loss: 0.5131 Iter 8857, loss: 0.5132 Iter 8858, loss: 0.5574 Iter 8859, loss: 0.5213 Iter 8860, loss: 0.5800 Iter 8861, loss: 0.5504 Iter 8862, loss: 0.5425 Iter 8863, loss: 0.5046 Iter 8864, loss: 0.5537 Iter 8865, loss: 0.5002 Iter 8866, loss: 0.5169 Iter 8867, loss: 0.5257 Iter 8868, loss: 0.5248 Iter 8869, loss: 0.5001 Iter 8870, loss: 0.5388 Iter 8871, loss: 0.5414 Iter 8872, loss: 0.5349 Iter 8873, loss: 0.5820 Iter 8874, loss: 0.5403 Iter 8875, loss: 0.5249 Iter 8876, loss: 0.5338 Iter 8877, loss: 0.5743 Iter 8878, loss: 0.5378 Iter 8879, loss: 0.4984 Iter 8880, loss: 0.5364 Iter 8881, loss: 0.5499 Iter 8882, loss: 0.5713 Iter 8883, loss: 0.5763 Iter 8884, loss: 0.5143 Iter 8885, loss: 0.5768

Iter 8886, loss: 0.5083 Iter 8887, loss: 0.5334 Iter 8888, loss: 0.5595 Iter 8889, loss: 0.5659 Iter 8890, loss: 0.4966 Iter 8891, loss: 0.5255 Iter 8892, loss: 0.5389 Iter 8893, loss: 0.5461 Iter 8894, loss: 0.5462 Iter 8895, loss: 0.4962 Iter 8896, loss: 0.5388 Iter 8897, loss: 0.5379 Iter 8898, loss: 0.5776 Iter 8899, loss: 0.5254 Iter 8900, loss: 0.5472 Iter 8901, loss: 0.4896 Iter 8902, loss: 0.5516 Iter 8903, loss: 0.4793 Iter 8904, loss: 0.5473 Iter 8905, loss: 0.5308 Iter 8906, loss: 0.5294 Iter 8907, loss: 0.5496 Iter 8908, loss: 0.5357 Iter 8909, loss: 0.5052 Iter 8910, loss: 0.5470 Iter 8911, loss: 0.5046 Iter 8912, loss: 0.5808 Iter 8913, loss: 0.5168 Iter 8914, loss: 0.5025 Iter 8915, loss: 0.5397 Iter 8916, loss: 0.5568 Iter 8917, loss: 0.5321 Iter 8918, loss: 0.5606 Iter 8919, loss: 0.5530 Iter 8920, loss: 0.6004 Iter 8921, loss: 0.4973 Iter 8922, loss: 0.5341 Iter 8923, loss: 0.5526 Iter 8924, loss: 0.5217 Iter 8925, loss: 0.5657 Iter 8926, loss: 0.5736 Iter 8927, loss: 0.5330

Iter 8928, loss: 0.5426 Iter 8929, loss: 0.5485 Iter 8930, loss: 0.5260 Iter 8931, loss: 0.5470 Iter 8932, loss: 0.4804 Iter 8933, loss: 0.5010 Iter 8934, loss: 0.5407 Iter 8935, loss: 0.5344 Iter 8936, loss: 0.5014 Iter 8937, loss: 0.5225 Iter 8938, loss: 0.4785 Iter 8939, loss: 0.5113 Iter 8940, loss: 0.5275 Iter 8941, loss: 0.5211 Iter 8942, loss: 0.5023 Iter 8943, loss: 0.5815 Iter 8944, loss: 0.4977 Iter 8945, loss: 0.4982 Iter 8946, loss: 0.5557 Iter 8947, loss: 0.6033 Iter 8948, loss: 0.5342 Iter 8949, loss: 0.5749 Iter 8950, loss: 0.5035 Iter 8951, loss: 0.5156 Iter 8952, loss: 0.5619 Iter 8953, loss: 0.5306 Iter 8954, loss: 0.5195 Iter 8955, loss: 0.5300 Iter 8956, loss: 0.5306 Iter 8957, loss: 0.5396 Iter 8958, loss: 0.5612 Iter 8959, loss: 0.5019 Iter 8960, loss: 0.5582 Iter 8961, loss: 0.5112 Iter 8962, loss: 0.5305 Iter 8963, loss: 0.5146 Iter 8964, loss: 0.5949 Iter 8965, loss: 0.5714 Iter 8966, loss: 0.5583 Iter 8967, loss: 0.5103 Iter 8968, loss: 0.4847 Iter 8969, loss: 0.5090

Iter 8970, loss: 0.5460 Iter 8971, loss: 0.5025 Iter 8972, loss: 0.5466 Iter 8973, loss: 0.5128 Iter 8974, loss: 0.5123 Iter 8975, loss: 0.5127 Iter 8976, loss: 0.5989 Iter 8977, loss: 0.5132 Iter 8978, loss: 0.5641 Iter 8979, loss: 0.5585 Iter 8980, loss: 0.4886 Iter 8981, loss: 0.5475 Iter 8982, loss: 0.5620 Iter 8983, loss: 0.4981 Iter 8984, loss: 0.5483 Iter 8985, loss: 0.5195 Iter 8986, loss: 0.5094 Iter 8987, loss: 0.5602 Iter 8988, loss: 0.5069 Iter 8989, loss: 0.5148 Iter 8990, loss: 0.5437 Iter 8991, loss: 0.5380 Iter 8992, loss: 0.5739 Iter 8993, loss: 0.5408 Iter 8994, loss: 0.5124 Iter 8995, loss: 0.5184 Iter 8996, loss: 0.5504 Iter 8997, loss: 0.4922 Iter 8998, loss: 0.5400 Iter 8999, loss: 0.5457 Iter 9000, loss: 0.5340 Iter 9001, loss: 0.5504 Iter 9002, loss: 0.5155 Iter 9003, loss: 0.5401 Iter 9004, loss: 0.5357 Iter 9005, loss: 0.5097 Iter 9006, loss: 0.5488 Iter 9007, loss: 0.4856 Iter 9008, loss: 0.4960 Iter 9009, loss: 0.5271 Iter 9010, loss: 0.5129 Iter 9011, loss: 0.5112

Iter 9012, loss: 0.4914 Iter 9013, loss: 0.5486 Iter 9014, loss: 0.5496 Iter 9015, loss: 0.5114 Iter 9016, loss: 0.5589 Iter 9017, loss: 0.5173 Iter 9018, loss: 0.5161 Iter 9019, loss: 0.4930 Iter 9020, loss: 0.5645 Iter 9021, loss: 0.5475 Iter 9022, loss: 0.5294 Iter 9023, loss: 0.5557 Iter 9024, loss: 0.5244 Iter 9025, loss: 0.5356 Iter 9026, loss: 0.5100 Iter 9027, loss: 0.5269 Iter 9028, loss: 0.5321 Iter 9029, loss: 0.5393 Iter 9030, loss: 0.5358 Iter 9031, loss: 0.5475 Iter 9032, loss: 0.4874 Iter 9033, loss: 0.5862 Iter 9034, loss: 0.5377 Iter 9035, loss: 0.6225 Iter 9036, loss: 0.5370 Iter 9037, loss: 0.5536 Iter 9038, loss: 0.5056 Iter 9039, loss: 0.5190 Iter 9040, loss: 0.5814 Iter 9041, loss: 0.5241 Iter 9042, loss: 0.4695 Iter 9043, loss: 0.5322 Iter 9044, loss: 0.5368 Iter 9045, loss: 0.5449 Iter 9046, loss: 0.4854 Iter 9047, loss: 0.5668 Iter 9048, loss: 0.5118 Iter 9049, loss: 0.5072 Iter 9050, loss: 0.5479 Iter 9051, loss: 0.5135 Iter 9052, loss: 0.5089 Iter 9053, loss: 0.5095

Iter 9054, loss: 0.5524 Iter 9055, loss: 0.5228 Iter 9056, loss: 0.5642 Iter 9057, loss: 0.5241 Iter 9058, loss: 0.5530 Iter 9059, loss: 0.5691 Iter 9060, loss: 0.5162 Iter 9061, loss: 0.5021 Iter 9062, loss: 0.5207 Iter 9063, loss: 0.5231 Iter 9064, loss: 0.5244 Iter 9065, loss: 0.5013 Iter 9066, loss: 0.5334 Iter 9067, loss: 0.5391 Iter 9068, loss: 0.5452 Iter 9069, loss: 0.5298 Iter 9070, loss: 0.5336 Iter 9071, loss: 0.5603 Iter 9072, loss: 0.5460 Iter 9073, loss: 0.5101 Iter 9074, loss: 0.5455 Iter 9075, loss: 0.5577 Iter 9076, loss: 0.5134 Iter 9077, loss: 0.5350 Iter 9078, loss: 0.5184 Iter 9079, loss: 0.5480 Iter 9080, loss: 0.5723 Iter 9081, loss: 0.5663 Iter 9082, loss: 0.5569 Iter 9083, loss: 0.5249 Iter 9084, loss: 0.5310 Iter 9085, loss: 0.4784 Iter 9086, loss: 0.5202 Iter 9087, loss: 0.5543 Iter 9088, loss: 0.5572 Iter 9089, loss: 0.5132 Iter 9090, loss: 0.4964 Iter 9091, loss: 0.5235 Iter 9092, loss: 0.5595 Iter 9093, loss: 0.5168 Iter 9094, loss: 0.4915

Iter 9095, loss: 0.5316 Iter 9096, loss: 0.5335 Iter 9097, loss: 0.5302 Iter 9098, loss: 0.5510 Iter 9099, loss: 0.5373 Iter 9100, loss: 0.6129 Iter 9101, loss: 0.5080 Iter 9102, loss: 0.4989 Iter 9103, loss: 0.4971 Iter 9104, loss: 0.5328 Iter 9105, loss: 0.5798 Iter 9106, loss: 0.5039 Iter 9107, loss: 0.5420 Iter 9108, loss: 0.5355 Iter 9109, loss: 0.5112 Iter 9110, loss: 0.5709 Iter 9111, loss: 0.5847 Iter 9112, loss: 0.4905 Iter 9113, loss: 0.6109 Iter 9114, loss: 0.5123 Iter 9115, loss: 0.5138 Iter 9116, loss: 0.5299 Iter 9117, loss: 0.5298 Iter 9118, loss: 0.5098 Iter 9119, loss: 0.5727 Iter 9120, loss: 0.4932 Iter 9121, loss: 0.5040 Iter 9122, loss: 0.5003 Iter 9123, loss: 0.5134 Iter 9124, loss: 0.5626 Iter 9125, loss: 0.5370 Iter 9126, loss: 0.5595 Iter 9127, loss: 0.5171 Iter 9128, loss: 0.5763 Iter 9129, loss: 0.5775 Iter 9130, loss: 0.4972 Iter 9131, loss: 0.5268 Iter 9132, loss: 0.4921 Iter 9133, loss: 0.4947 Iter 9134, loss: 0.5184 Iter 9135, loss: 0.5913 Iter 9136, loss: 0.5260

Iter 9137, loss: 0.5450 Iter 9138, loss: 0.5089 Iter 9139, loss: 0.5253 Iter 9140, loss: 0.5158 Iter 9141, loss: 0.5035 Iter 9142, loss: 0.5467 Iter 9143, loss: 0.5477 Iter 9144, loss: 0.5246 Iter 9145, loss: 0.5186 Iter 9146, loss: 0.5205 Iter 9147, loss: 0.5265 Iter 9148, loss: 0.5786 Iter 9149, loss: 0.5450 Iter 9150, loss: 0.4960 Iter 9151, loss: 0.5464 Iter 9152, loss: 0.5362 Iter 9153, loss: 0.5293 Iter 9154, loss: 0.4732 Iter 9155, loss: 0.5489 Iter 9156, loss: 0.5414 Iter 9157, loss: 0.4791 Iter 9158, loss: 0.5298 Iter 9159, loss: 0.5282 Iter 9160, loss: 0.5508 Iter 9161, loss: 0.5290 Iter 9162, loss: 0.5832 Iter 9163, loss: 0.4941 Iter 9164, loss: 0.5145 Iter 9165, loss: 0.5444 Iter 9166, loss: 0.5359 Iter 9167, loss: 0.5510 Iter 9168, loss: 0.5419 Iter 9169, loss: 0.5288 Iter 9170, loss: 0.5427 Iter 9171, loss: 0.5012 Iter 9172, loss: 0.5307 Iter 9173, loss: 0.5612 Iter 9174, loss: 0.5753 Iter 9175, loss: 0.5007 Iter 9176, loss: 0.5129 Iter 9177, loss: 0.5083 Iter 9178, loss: 0.5224

Iter 9179, loss: 0.5397 Iter 9180, loss: 0.5095 Iter 9181, loss: 0.4996 Iter 9182, loss: 0.5396 Iter 9183, loss: 0.5218 Iter 9184, loss: 0.5143 Iter 9185, loss: 0.5118 Iter 9186, loss: 0.5280 Iter 9187, loss: 0.5437 Iter 9188, loss: 0.5219 Iter 9189, loss: 0.5479 Iter 9190, loss: 0.5087 Iter 9191, loss: 0.5184 Iter 9192, loss: 0.5558 Iter 9193, loss: 0.5356 Iter 9194, loss: 0.5405 Iter 9195, loss: 0.5455 Iter 9196, loss: 0.4941 Iter 9197, loss: 0.5411 Iter 9198, loss: 0.5112 Iter 9199, loss: 0.5263 Iter 9200, loss: 0.5387 Iter 9201, loss: 0.4986 Iter 9202, loss: 0.5110 Iter 9203, loss: 0.5393 Iter 9204, loss: 0.5319 Iter 9205, loss: 0.5246 Iter 9206, loss: 0.5199 Iter 9207, loss: 0.5366 Iter 9208, loss: 0.5571 Iter 9209, loss: 0.5230 Iter 9210, loss: 0.5168 Iter 9211, loss: 0.5207 Iter 9212, loss: 0.5181 Iter 9213, loss: 0.5397 Iter 9214, loss: 0.5063 Iter 9215, loss: 0.5248 Iter 9216, loss: 0.5227 Iter 9217, loss: 0.5445 Iter 9218, loss: 0.5245 Iter 9219, loss: 0.4811 Iter 9220, loss: 0.5018

Iter 9221, loss: 0.5185 Iter 9222, loss: 0.5313 Iter 9223, loss: 0.5243 Iter 9224, loss: 0.5537 Iter 9225, loss: 0.5209 Iter 9226, loss: 0.5427 Iter 9227, loss: 0.5041 Iter 9228, loss: 0.5460 Iter 9229, loss: 0.5183 Iter 9230, loss: 0.5408 Iter 9231, loss: 0.5107 Iter 9232, loss: 0.4884 Iter 9233, loss: 0.5339 Iter 9234, loss: 0.5310 Iter 9235, loss: 0.5311 Iter 9236, loss: 0.5804 Iter 9237, loss: 0.5348 Iter 9238, loss: 0.5107 Iter 9239, loss: 0.5175 Iter 9240, loss: 0.5073 Iter 9241, loss: 0.5670 Iter 9242, loss: 0.5311 Iter 9243, loss: 0.5184 Iter 9244, loss: 0.5468 Iter 9245, loss: 0.5558 Iter 9246, loss: 0.5333 Iter 9247, loss: 0.5235 Iter 9248, loss: 0.5144 Iter 9249, loss: 0.5206 Iter 9250, loss: 0.5194 Iter 9251, loss: 0.5269 Iter 9252, loss: 0.5554 Iter 9253, loss: 0.5073 Iter 9254, loss: 0.5417 Iter 9255, loss: 0.5371 Iter 9256, loss: 0.4971 Iter 9257, loss: 0.5426 Iter 9258, loss: 0.5465 Iter 9259, loss: 0.5274 Iter 9260, loss: 0.5584 Iter 9261, loss: 0.5582

Iter 9262, loss: 0.5330 Iter 9263, loss: 0.5351 Iter 9264, loss: 0.5369 Iter 9265, loss: 0.5571 Iter 9266, loss: 0.5421 Iter 9267, loss: 0.5517 Iter 9268, loss: 0.4982 Iter 9269, loss: 0.5195 Iter 9270, loss: 0.5029 Iter 9271, loss: 0.5165 Iter 9272, loss: 0.5250 Iter 9273, loss: 0.5150 Iter 9274, loss: 0.5347 Iter 9275, loss: 0.5299 Iter 9276, loss: 0.5078 Iter 9277, loss: 0.5254 Iter 9278, loss: 0.4932 Iter 9279, loss: 0.5069 Iter 9280, loss: 0.4819 Iter 9281, loss: 0.5540 Iter 9282, loss: 0.5350 Iter 9283, loss: 0.5694 Iter 9284, loss: 0.5689 Iter 9285, loss: 0.5683 Iter 9286, loss: 0.5673 Iter 9287, loss: 0.5386 Iter 9288, loss: 0.5193 Iter 9289, loss: 0.5308 Iter 9290, loss: 0.5162 Iter 9291, loss: 0.4978 Iter 9292, loss: 0.5148 Iter 9293, loss: 0.5810 Iter 9294, loss: 0.5683 Iter 9295, loss: 0.4973 Iter 9296, loss: 0.5229 Iter 9297, loss: 0.5468 Iter 9298, loss: 0.5113 Iter 9299, loss: 0.4892 Iter 9300, loss: 0.5356 Iter 9301, loss: 0.4930 Iter 9302, loss: 0.5416

Iter 9303, loss: 0.5614 Iter 9304, loss: 0.5504 Iter 9305, loss: 0.5201 Iter 9306, loss: 0.5687 Iter 9307, loss: 0.5340 Iter 9308, loss: 0.5828 Iter 9309, loss: 0.4994 Iter 9310, loss: 0.5175 Iter 9311, loss: 0.5295 Iter 9312, loss: 0.5083 Iter 9313, loss: 0.5346 Iter 9314, loss: 0.5305 Iter 9315, loss: 0.5490 Iter 9316, loss: 0.4972 Iter 9317, loss: 0.4734 Iter 9318, loss: 0.5458 Iter 9319, loss: 0.5498 Iter 9320, loss: 0.5190 Iter 9321, loss: 0.5226 Iter 9322, loss: 0.5164 Iter 9323, loss: 0.5321 Iter 9324, loss: 0.5625 Iter 9325, loss: 0.5423 Iter 9326, loss: 0.5313 Iter 9327, loss: 0.5320 Iter 9328, loss: 0.5357 Iter 9329, loss: 0.4870 Iter 9330, loss: 0.5061 Iter 9331, loss: 0.5495 Iter 9332, loss: 0.5124 Iter 9333, loss: 0.5383 Iter 9334, loss: 0.4966 Iter 9335, loss: 0.5145 Iter 9336, loss: 0.5457 Iter 9337, loss: 0.5574 Iter 9338, loss: 0.5112 Iter 9339, loss: 0.5277 Iter 9340, loss: 0.5238 Iter 9341, loss: 0.5221 Iter 9342, loss: 0.5467 Iter 9343, loss: 0.5429 Iter 9344, loss: 0.5241

Iter 9345, loss: 0.4922 Iter 9346, loss: 0.5397 Iter 9347, loss: 0.5156 Iter 9348, loss: 0.4897 Iter 9349, loss: 0.5257 Iter 9350, loss: 0.5584 Iter 9351, loss: 0.5297 Iter 9352, loss: 0.5720 Iter 9353, loss: 0.5509 Iter 9354, loss: 0.5434 Iter 9355, loss: 0.5164 Iter 9356, loss: 0.5272 Iter 9357, loss: 0.5406 Iter 9358, loss: 0.5007 Iter 9359, loss: 0.5100 Iter 9360, loss: 0.5091 Iter 9361, loss: 0.5065 Iter 9362, loss: 0.5094 Iter 9363, loss: 0.5744 Iter 9364, loss: 0.4768 Iter 9365, loss: 0.5468 Iter 9366, loss: 0.5739 Iter 9367, loss: 0.5195 Iter 9368, loss: 0.5330 Iter 9369, loss: 0.5237 Iter 9370, loss: 0.5515 Iter 9371, loss: 0.5151 Iter 9372, loss: 0.5122 Iter 9373, loss: 0.5783 Iter 9374, loss: 0.4997 Iter 9375, loss: 0.5389 Iter 9376, loss: 0.4951 Iter 9377, loss: 0.5269 Iter 9378, loss: 0.5568 Iter 9379, loss: 0.5303 Iter 9380, loss: 0.5302 Iter 9381, loss: 0.5083 Iter 9382, loss: 0.5149 Iter 9383, loss: 0.5165 Iter 9384, loss: 0.5171 Iter 9385, loss: 0.5265

Iter 9386, loss: 0.5041 Iter 9387, loss: 0.5582 Iter 9388, loss: 0.5271 Iter 9389, loss: 0.5182 Iter 9390, loss: 0.5219 Iter 9391, loss: 0.4664 Iter 9392, loss: 0.5303 Iter 9393, loss: 0.5344 Iter 9394, loss: 0.5311 Iter 9395, loss: 0.5232 Iter 9396, loss: 0.4987 Iter 9397, loss: 0.5591 Iter 9398, loss: 0.5468 Iter 9399, loss: 0.5096 Iter 9400, loss: 0.5470 Iter 9401, loss: 0.5446 Iter 9402, loss: 0.5087 Iter 9403, loss: 0.4972 Iter 9404, loss: 0.5337 Iter 9405, loss: 0.4814 Iter 9406, loss: 0.5158 Iter 9407, loss: 0.5641 Iter 9408, loss: 0.5086 Iter 9409, loss: 0.5425 Iter 9410, loss: 0.5425 Iter 9411, loss: 0.5804 Iter 9412, loss: 0.5555 Iter 9413, loss: 0.5636 Iter 9414, loss: 0.5327 Iter 9415, loss: 0.5310 Iter 9416, loss: 0.5653 Iter 9417, loss: 0.4922 Iter 9418, loss: 0.5190 Iter 9419, loss: 0.5000 Iter 9420, loss: 0.4976 Iter 9421, loss: 0.5040 Iter 9422, loss: 0.5371 Iter 9423, loss: 0.5144 Iter 9424, loss: 0.5419 Iter 9425, loss: 0.5259 Iter 9426, loss: 0.5245

Iter 9427, loss: 0.5093 Iter 9428, loss: 0.5095 Iter 9429, loss: 0.5205 Iter 9430, loss: 0.5502 Iter 9431, loss: 0.5423 Iter 9432, loss: 0.5286 Iter 9433, loss: 0.5509 Iter 9434, loss: 0.5303 Iter 9435, loss: 0.5131 Iter 9436, loss: 0.5435 Iter 9437, loss: 0.5441 Iter 9438, loss: 0.4888 Iter 9439, loss: 0.5616 Iter 9440, loss: 0.5264 Iter 9441, loss: 0.5483 Iter 9442, loss: 0.4780 Iter 9443, loss: 0.5284 Iter 9444, loss: 0.5293 Iter 9445, loss: 0.5534 Iter 9446, loss: 0.5358 Iter 9447, loss: 0.5662 Iter 9448, loss: 0.4998 Iter 9449, loss: 0.5192 Iter 9450, loss: 0.4970 Iter 9451, loss: 0.4962 Iter 9452, loss: 0.5316 Iter 9453, loss: 0.5154 Iter 9454, loss: 0.5269 Iter 9455, loss: 0.5646 Iter 9456, loss: 0.5299 Iter 9457, loss: 0.5596 Iter 9458, loss: 0.5406 Iter 9459, loss: 0.5559 Iter 9460, loss: 0.5600 Iter 9461, loss: 0.5481 Iter 9462, loss: 0.5332 Iter 9463, loss: 0.4840 Iter 9464, loss: 0.5114 Iter 9465, loss: 0.5368 Iter 9466, loss: 0.5265 Iter 9467, loss: 0.5273

Iter 9468, loss: 0.5188 Iter 9469, loss: 0.5195 Iter 9470, loss: 0.4761 Iter 9471, loss: 0.5259 Iter 9472, loss: 0.5243 Iter 9473, loss: 0.5519 Iter 9474, loss: 0.5840 Iter 9475, loss: 0.4973 Iter 9476, loss: 0.5154 Iter 9477, loss: 0.5376 Iter 9478, loss: 0.5267 Iter 9479, loss: 0.5122 Iter 9480, loss: 0.4908 Iter 9481, loss: 0.5086 Iter 9482, loss: 0.5090 Iter 9483, loss: 0.5260 Iter 9484, loss: 0.5232 Iter 9485, loss: 0.5124 Iter 9486, loss: 0.4786 Iter 9487, loss: 0.4979 Iter 9488, loss: 0.5731 Iter 9489, loss: 0.5429 Iter 9490, loss: 0.5251 Iter 9491, loss: 0.5157 Iter 9492, loss: 0.5266 Iter 9493, loss: 0.5016 Iter 9494, loss: 0.5097 Iter 9495, loss: 0.5169 Iter 9496, loss: 0.5030 Iter 9497, loss: 0.5399 Iter 9498, loss: 0.5317 Iter 9499, loss: 0.5222 Iter 9500, loss: 0.5279 Iter 9501, loss: 0.5323 Iter 9502, loss: 0.5316 Iter 9503, loss: 0.4890 Iter 9504, loss: 0.5431 Iter 9505, loss: 0.5107 Iter 9506, loss: 0.5205 Iter 9507, loss: 0.5284 Iter 9508, loss: 0.5384 Iter 9509, loss: 0.4954

Iter 9510, loss: 0.5578 Iter 9511, loss: 0.5316 Iter 9512, loss: 0.5187 Iter 9513, loss: 0.4748 Iter 9514, loss: 0.5207 Iter 9515, loss: 0.5173 Iter 9516, loss: 0.5030 Iter 9517, loss: 0.4957 Iter 9518, loss: 0.5163 Iter 9519, loss: 0.5135 Iter 9520, loss: 0.5072 Iter 9521, loss: 0.5073 Iter 9522, loss: 0.5555 Iter 9523, loss: 0.4925 Iter 9524, loss: 0.4995 Iter 9525, loss: 0.5401 Iter 9526, loss: 0.5320 Iter 9527, loss: 0.4979 Iter 9528, loss: 0.5379 Iter 9529, loss: 0.5115 Iter 9530, loss: 0.5010 Iter 9531, loss: 0.5494 Iter 9532, loss: 0.5310 Iter 9533, loss: 0.5343 Iter 9534, loss: 0.5295 Iter 9535, loss: 0.5311 Iter 9536, loss: 0.5236 Iter 9537, loss: 0.5470 Iter 9538, loss: 0.5708 Iter 9539, loss: 0.5316 Iter 9540, loss: 0.5058 Iter 9541, loss: 0.5447 Iter 9542, loss: 0.5336 Iter 9543, loss: 0.5266 Iter 9544, loss: 0.5072 Iter 9545, loss: 0.5046 Iter 9546, loss: 0.5198 Iter 9547, loss: 0.5069 Iter 9548, loss: 0.5237 Iter 9549, loss: 0.5388 Iter 9550, loss: 0.5504

Iter 9551, loss: 0.5436 Iter 9552, loss: 0.5474 Iter 9553, loss: 0.5173 Iter 9554, loss: 0.5314 Iter 9555, loss: 0.5097 Iter 9556, loss: 0.5334 Iter 9557, loss: 0.5369 Iter 9558, loss: 0.5391 Iter 9559, loss: 0.5216 Iter 9560, loss: 0.4960 Iter 9561, loss: 0.5348 Iter 9562, loss: 0.5499 Iter 9563, loss: 0.5426 Iter 9564, loss: 0.4889 Iter 9565, loss: 0.5267 Iter 9566, loss: 0.5259 Iter 9567, loss: 0.5572 Iter 9568, loss: 0.4977 Iter 9569, loss: 0.5054 Iter 9570, loss: 0.5189 Iter 9571, loss: 0.5375 Iter 9572, loss: 0.5270 Iter 9573, loss: 0.5315 Iter 9574, loss: 0.5152 Iter 9575, loss: 0.5218 Iter 9576, loss: 0.5001 Iter 9577, loss: 0.5156 Iter 9578, loss: 0.5108 Iter 9579, loss: 0.5637 Iter 9580, loss: 0.5577 Iter 9581, loss: 0.5218 Iter 9582, loss: 0.5577 Iter 9583, loss: 0.5187 Iter 9584, loss: 0.5287 Iter 9585, loss: 0.5363 Iter 9586, loss: 0.5054 Iter 9587, loss: 0.4985 Iter 9588, loss: 0.5141 Iter 9589, loss: 0.5100 Iter 9590, loss: 0.5446 Iter 9591, loss: 0.5483 Iter 9592, loss: 0.5382

Iter 9593, loss: 0.5149 Iter 9594, loss: 0.5293 Iter 9595, loss: 0.5334 Iter 9596, loss: 0.5037 Iter 9597, loss: 0.5105 Iter 9598, loss: 0.5211 Iter 9599, loss: 0.5158 Iter 9600, loss: 0.5513 Iter 9601, loss: 0.4984 Iter 9602, loss: 0.5097 Iter 9603, loss: 0.5103 Iter 9604, loss: 0.5534 Iter 9605, loss: 0.4790 Iter 9606, loss: 0.5193 Iter 9607, loss: 0.5170 Iter 9608, loss: 0.5055 Iter 9609, loss: 0.5195 Iter 9610, loss: 0.5327 Iter 9611, loss: 0.5369 Iter 9612, loss: 0.5009 Iter 9613, loss: 0.5040 Iter 9614, loss: 0.5474 Iter 9615, loss: 0.5546 Iter 9616, loss: 0.5500 Iter 9617, loss: 0.5366 Iter 9618, loss: 0.5553 Iter 9619, loss: 0.4920 Iter 9620, loss: 0.5411 Iter 9621, loss: 0.4831 Iter 9622, loss: 0.5053 Iter 9623, loss: 0.4974 Iter 9624, loss: 0.4859 Iter 9625, loss: 0.5360 Iter 9626, loss: 0.5176 Iter 9627, loss: 0.4738 Iter 9628, loss: 0.4942 Iter 9629, loss: 0.5507 Iter 9630, loss: 0.4952 Iter 9631, loss: 0.5231 Iter 9632, loss: 0.5045 Iter 9633, loss: 0.5612 Iter 9634, loss: 0.5194

Iter 9635, loss: 0.5466 Iter 9636, loss: 0.5262 Iter 9637, loss: 0.4990 Iter 9638, loss: 0.5624 Iter 9639, loss: 0.5390 Iter 9640, loss: 0.5281 Iter 9641, loss: 0.4832 Iter 9642, loss: 0.5492 Iter 9643, loss: 0.4958 Iter 9644, loss: 0.5629 Iter 9645, loss: 0.5703 Iter 9646, loss: 0.5081 Iter 9647, loss: 0.5441 Iter 9648, loss: 0.5634 Iter 9649, loss: 0.5748 Iter 9650, loss: 0.5523 Iter 9651, loss: 0.5492 Iter 9652, loss: 0.5745 Iter 9653, loss: 0.5087 Iter 9654, loss: 0.5513 Iter 9655, loss: 0.5060 Iter 9656, loss: 0.5439 Iter 9657, loss: 0.5219 Iter 9658, loss: 0.5260 Iter 9659, loss: 0.5100 Iter 9660, loss: 0.5289 Iter 9661, loss: 0.5497 Iter 9662, loss: 0.5427 Iter 9663, loss: 0.5116 Iter 9664, loss: 0.5030 Iter 9665, loss: 0.5774 Iter 9666, loss: 0.5236 Iter 9667, loss: 0.4656 Iter 9668, loss: 0.5315 Iter 9669, loss: 0.5070 Iter 9670, loss: 0.5389 Iter 9671, loss: 0.5584 Iter 9672, loss: 0.5483 Iter 9673, loss: 0.5473 Iter 9674, loss: 0.5313 Iter 9675, loss: 0.5491 Iter 9676, loss: 0.5024

Iter 9677, loss: 0.5469 Iter 9678, loss: 0.5000 Iter 9679, loss: 0.4667 Iter 9680, loss: 0.4892 Iter 9681, loss: 0.5212 Iter 9682, loss: 0.5687 Iter 9683, loss: 0.5502 Iter 9684, loss: 0.5074 Iter 9685, loss: 0.5258 Iter 9686, loss: 0.4756 Iter 9687, loss: 0.5298 Iter 9688, loss: 0.5767 Iter 9689, loss: 0.5132 Iter 9690, loss: 0.4856 Iter 9691, loss: 0.5565 Iter 9692, loss: 0.5414 Iter 9693, loss: 0.5373 Iter 9694, loss: 0.4919 Iter 9695, loss: 0.5170 Iter 9696, loss: 0.5085 Iter 9697, loss: 0.5336 Iter 9698, loss: 0.5211 Iter 9699, loss: 0.4908 Iter 9700, loss: 0.4987 Iter 9701, loss: 0.5277 Iter 9702, loss: 0.5267 Iter 9703, loss: 0.5347 Iter 9704, loss: 0.5036 Iter 9705, loss: 0.4646 Iter 9706, loss: 0.5445 Iter 9707, loss: 0.5509 Iter 9708, loss: 0.5114 Iter 9709, loss: 0.5307 Iter 9710, loss: 0.5332 Iter 9711, loss: 0.5292 Iter 9712, loss: 0.5445 Iter 9713, loss: 0.5515 Iter 9714, loss: 0.5174 Iter 9715, loss: 0.5167 Iter 9716, loss: 0.4796 Iter 9717, loss: 0.5581 Iter 9718, loss: 0.5081

Iter 9719, loss: 0.5185 Iter 9720, loss: 0.4979 Iter 9721, loss: 0.5298 Iter 9722, loss: 0.5600 Iter 9723, loss: 0.5267 Iter 9724, loss: 0.5294 Iter 9725, loss: 0.5057 Iter 9726, loss: 0.5503 Iter 9727, loss: 0.5287 Iter 9728, loss: 0.5033 Iter 9729, loss: 0.5081 Iter 9730, loss: 0.5170 Iter 9731, loss: 0.5033 Iter 9732, loss: 0.5103 Iter 9733, loss: 0.5047 Iter 9734, loss: 0.5096 Iter 9735, loss: 0.5277 Iter 9736, loss: 0.4804 Iter 9737, loss: 0.4695 Iter 9738, loss: 0.5136 Iter 9739, loss: 0.5663 Iter 9740, loss: 0.5205 Iter 9741, loss: 0.5706 Iter 9742, loss: 0.5471 Iter 9743, loss: 0.5000 Iter 9744, loss: 0.5469 Iter 9745, loss: 0.5250 Iter 9746, loss: 0.4999 Iter 9747, loss: 0.5428 Iter 9748, loss: 0.5342 Iter 9749, loss: 0.5227 Iter 9750, loss: 0.5092 Iter 9751, loss: 0.4923 Iter 9752, loss: 0.5249 Iter 9753, loss: 0.5373 Iter 9754, loss: 0.4916 Iter 9755, loss: 0.5158 Iter 9756, loss: 0.4873 Iter 9757, loss: 0.5008 Iter 9758, loss: 0.4933 Iter 9759, loss: 0.5141 Iter 9760, loss: 0.5275

Iter 9761, loss: 0.5407 Iter 9762, loss: 0.5099 Iter 9763, loss: 0.5505 Iter 9764, loss: 0.5321 Iter 9765, loss: 0.5199 Iter 9766, loss: 0.5232 Iter 9767, loss: 0.4842 Iter 9768, loss: 0.5527 Iter 9769, loss: 0.5270 Iter 9770, loss: 0.5038 Iter 9771, loss: 0.5603 Iter 9772, loss: 0.5156 Iter 9773, loss: 0.5477 Iter 9774, loss: 0.4973 Iter 9775, loss: 0.5619 Iter 9776, loss: 0.5115 Iter 9777, loss: 0.5323 Iter 9778, loss: 0.5282 Iter 9779, loss: 0.5248 Iter 9780, loss: 0.5371 Iter 9781, loss: 0.5156 Iter 9782, loss: 0.5214 Iter 9783, loss: 0.5159 Iter 9784, loss: 0.5565 Iter 9785, loss: 0.5359 Iter 9786, loss: 0.5695 Iter 9787, loss: 0.5824 Iter 9788, loss: 0.5585 Iter 9789, loss: 0.5513 Iter 9790, loss: 0.5046 Iter 9791, loss: 0.4900 Iter 9792, loss: 0.5589 Iter 9793, loss: 0.5232 Iter 9794, loss: 0.5106 Iter 9795, loss: 0.5339 Iter 9796, loss: 0.4796 Iter 9797, loss: 0.5448 Iter 9798, loss: 0.5110 Iter 9799, loss: 0.4823 Iter 9800, loss: 0.5464 Iter 9801, loss: 0.5252 Iter 9802, loss: 0.5254

Iter 9803, loss: 0.5094 Iter 9804, loss: 0.5023 Iter 9805, loss: 0.4677 Iter 9806, loss: 0.5465 Iter 9807, loss: 0.5326 Iter 9808, loss: 0.5199 Iter 9809, loss: 0.5097 Iter 9810, loss: 0.5279 Iter 9811, loss: 0.5029 Iter 9812, loss: 0.5669 Iter 9813, loss: 0.4792 Iter 9814, loss: 0.5089 Iter 9815, loss: 0.5535 Iter 9816, loss: 0.5033 Iter 9817, loss: 0.4955 Iter 9818, loss: 0.5006 Iter 9819, loss: 0.5055 Iter 9820, loss: 0.5328 Iter 9821, loss: 0.4951 Iter 9822, loss: 0.5720 Iter 9823, loss: 0.5267 Iter 9824, loss: 0.5185 Iter 9825, loss: 0.5673 Iter 9826, loss: 0.5441 Iter 9827, loss: 0.5462 Iter 9828, loss: 0.5595 Iter 9829, loss: 0.5473 Iter 9830, loss: 0.5452 Iter 9831, loss: 0.4903 Iter 9832, loss: 0.5465 Iter 9833, loss: 0.5213 Iter 9834, loss: 0.4928 Iter 9835, loss: 0.5096 Iter 9836, loss: 0.5045 Iter 9837, loss: 0.5413 Iter 9838, loss: 0.5525 Iter 9839, loss: 0.5655 Iter 9840, loss: 0.5283 Iter 9841, loss: 0.4941 Iter 9842, loss: 0.4878 Iter 9843, loss: 0.5533 Iter 9844, loss: 0.5178

Iter 9845, loss: 0.5368 Iter 9846, loss: 0.4876 Iter 9847, loss: 0.4911 Iter 9848, loss: 0.5305 Iter 9849, loss: 0.4859 Iter 9850, loss: 0.5392 Iter 9851, loss: 0.5123 Iter 9852, loss: 0.4570 Iter 9853, loss: 0.5213 Iter 9854, loss: 0.5389 Iter 9855, loss: 0.5186 Iter 9856, loss: 0.4676 Iter 9857, loss: 0.5152 Iter 9858, loss: 0.5290 Iter 9859, loss: 0.5674 Iter 9860, loss: 0.4765 Iter 9861, loss: 0.5097 Iter 9862, loss: 0.5240 Iter 9863, loss: 0.5030 Iter 9864, loss: 0.5293 Iter 9865, loss: 0.4873 Iter 9866, loss: 0.4986 Iter 9867, loss: 0.5144 Iter 9868, loss: 0.5154 Iter 9869, loss: 0.5145 Iter 9870, loss: 0.5087 Iter 9871, loss: 0.5520 Iter 9872, loss: 0.4833 Iter 9873, loss: 0.5425 Iter 9874, loss: 0.5024 Iter 9875, loss: 0.5054 Iter 9876, loss: 0.4935 Iter 9877, loss: 0.4988 Iter 9878, loss: 0.5316 Iter 9879, loss: 0.5139 Iter 9880, loss: 0.5247 Iter 9881, loss: 0.4865 Iter 9882, loss: 0.5138 Iter 9883, loss: 0.5112 Iter 9884, loss: 0.5082 Iter 9885, loss: 0.5014 Iter 9886, loss: 0.5217

Iter 9887, loss: 0.5096 Iter 9888, loss: 0.5168 Iter 9889, loss: 0.5443 Iter 9890, loss: 0.5168 Iter 9891, loss: 0.4900 Iter 9892, loss: 0.5184 Iter 9893, loss: 0.4982 Iter 9894, loss: 0.4999 Iter 9895, loss: 0.5414 Iter 9896, loss: 0.4597 Iter 9897, loss: 0.5153 Iter 9898, loss: 0.5004 Iter 9899, loss: 0.5559 Iter 9900, loss: 0.4894 Iter 9901, loss: 0.5291 Iter 9902, loss: 0.4725 Iter 9903, loss: 0.4889 Iter 9904, loss: 0.4750 Iter 9905, loss: 0.5300 Iter 9906, loss: 0.5000 Iter 9907, loss: 0.5229 Iter 9908, loss: 0.5382 Iter 9909, loss: 0.5178 Iter 9910, loss: 0.5367 Iter 9911, loss: 0.5157 Iter 9912, loss: 0.5007 Iter 9913, loss: 0.5163 Iter 9914, loss: 0.5509 Iter 9915, loss: 0.5336 Iter 9916, loss: 0.5075 Iter 9917, loss: 0.5455 Iter 9918, loss: 0.4987 Iter 9919, loss: 0.5221 Iter 9920, loss: 0.5097 Iter 9921, loss: 0.5352 Iter 9922, loss: 0.5711 Iter 9923, loss: 0.4932 Iter 9924, loss: 0.5006 Iter 9925, loss: 0.5142 Iter 9926, loss: 0.5067 Iter 9927, loss: 0.5329 Iter 9928, loss: 0.5221

Iter 9929, loss: 0.4965 Iter 9930, loss: 0.5940 Iter 9931, loss: 0.5291 Iter 9932, loss: 0.5212 Iter 9933, loss: 0.5099 Iter 9934, loss: 0.4967 Iter 9935, loss: 0.5141 Iter 9936, loss: 0.4806 Iter 9937, loss: 0.5205 Iter 9938, loss: 0.5482 Iter 9939, loss: 0.4823 Iter 9940, loss: 0.4982 Iter 9941, loss: 0.5647 Iter 9942, loss: 0.4772 Iter 9943, loss: 0.5172 Iter 9944, loss: 0.5700 Iter 9945, loss: 0.4876 Iter 9946, loss: 0.5354 Iter 9947, loss: 0.4987 Iter 9948, loss: 0.5410 Iter 9949, loss: 0.5211 Iter 9950, loss: 0.4760 Iter 9951, loss: 0.5081 Iter 9952, loss: 0.4753 Iter 9953, loss: 0.4912 Iter 9954, loss: 0.5470 Iter 9955, loss: 0.4887 Iter 9956, loss: 0.5356 Iter 9957, loss: 0.5377 Iter 9958, loss: 0.4983 Iter 9959, loss: 0.5172 Iter 9960, loss: 0.5305 Iter 9961, loss: 0.5468 Iter 9962, loss: 0.5673 Iter 9963, loss: 0.5352 Iter 9964, loss: 0.5533 Iter 9965, loss: 0.5175 Iter 9966, loss: 0.5222 Iter 9967, loss: 0.5016 Iter 9968, loss: 0.5388 Iter 9969, loss: 0.5387 Iter 9970, loss: 0.5299

Iter 9971, loss: 0.5731 Iter 9972, loss: 0.5179 Iter 9973, loss: 0.5036 Iter 9974, loss: 0.5148 Iter 9975, loss: 0.5441 Iter 9976, loss: 0.5449 Iter 9977, loss: 0.5468 Iter 9978, loss: 0.5369 Iter 9979, loss: 0.4883 Iter 9980, loss: 0.5243 Iter 9981, loss: 0.5160 Iter 9982, loss: 0.5529 Iter 9983, loss: 0.5049 Iter 9984, loss: 0.4594 Iter 9985, loss: 0.5086 Iter 9986, loss: 0.4905 Iter 9987, loss: 0.5361 Iter 9988, loss: 0.4923 Iter 9989, loss: 0.5347 Iter 9990, loss: 0.5344 Iter 9991, loss: 0.4845 Iter 9992, loss: 0.5008 Iter 9993, loss: 0.5575 Iter 9994, loss: 0.4965 Iter 9995, loss: 0.5818 Iter 9996, loss: 0.5198 Iter 9997, loss: 0.5324 Iter 9998, loss: 0.4766 Iter 9999, loss: 0.4666 Iter 10000, loss: 0.4925 Iter 10001, loss: 0.4480 Iter 10002, loss: 0.4971 Iter 10003, loss: 0.5388 Iter 10004, loss: 0.5429 Iter 10005, loss: 0.5289 Iter 10006, loss: 0.5637 Iter 10007, loss: 0.4936 Iter 10008, loss: 0.5014 Iter 10009, loss: 0.5290 Iter 10010, loss: 0.5226 Iter 10011, loss: 0.4897 Iter 10012, loss: 0.5331

Iter 10013, loss: 0.5405 Iter 10014, loss: 0.5325 Iter 10015, loss: 0.4783 Iter 10016, loss: 0.5200 Iter 10017, loss: 0.5402 Iter 10018, loss: 0.4877 Iter 10019, loss: 0.5300 Iter 10020, loss: 0.5055 Iter 10021, loss: 0.5230 Iter 10022, loss: 0.5009 Iter 10023, loss: 0.4914 Iter 10024, loss: 0.5167 Iter 10025, loss: 0.5442 Iter 10026, loss: 0.4827 Iter 10027, loss: 0.5168 Iter 10028, loss: 0.4976 Iter 10029, loss: 0.5406 Iter 10030, loss: 0.5193 Iter 10031, loss: 0.5436 Iter 10032, loss: 0.5139 Iter 10033, loss: 0.5168 Iter 10034, loss: 0.5115 Iter 10035, loss: 0.5170 Iter 10036, loss: 0.5328 Iter 10037, loss: 0.5055 Iter 10038, loss: 0.5097 Iter 10039, loss: 0.5303 Iter 10040, loss: 0.5375 Iter 10041, loss: 0.5045 Iter 10042, loss: 0.4812 Iter 10043, loss: 0.5269 Iter 10044, loss: 0.5382 Iter 10045, loss: 0.5434 Iter 10046, loss: 0.5144 Iter 10047, loss: 0.5716 Iter 10048, loss: 0.5571 Iter 10049, loss: 0.4701 Iter 10050, loss: 0.5721 Iter 10051, loss: 0.5134 Iter 10052, loss: 0.5124 Iter 10053, loss: 0.5484 Iter 10054, loss: 0.5248

Iter 10055, loss: 0.5043 Iter 10056, loss: 0.5352 Iter 10057, loss: 0.5243 Iter 10058, loss: 0.5314 Iter 10059, loss: 0.4860 Iter 10060, loss: 0.5079 Iter 10061, loss: 0.5085 Iter 10062, loss: 0.5103 Iter 10063, loss: 0.5273 Iter 10064, loss: 0.5486 Iter 10065, loss: 0.5447 Iter 10066, loss: 0.5415 Iter 10067, loss: 0.5072 Iter 10068, loss: 0.5340 Iter 10069, loss: 0.5256 Iter 10070, loss: 0.5233 Iter 10071, loss: 0.5097 Iter 10072, loss: 0.5304 Iter 10073, loss: 0.4884 Iter 10074, loss: 0.5019 Iter 10075, loss: 0.4991 Iter 10076, loss: 0.5235 Iter 10077, loss: 0.4828 Iter 10078, loss: 0.5047 Iter 10079, loss: 0.5017 Iter 10080, loss: 0.5159 Iter 10081, loss: 0.5376 Iter 10082, loss: 0.4987 Iter 10083, loss: 0.5018 Iter 10084, loss: 0.5300 Iter 10085, loss: 0.4933 Iter 10086, loss: 0.5173 Iter 10087, loss: 0.4937 Iter 10088, loss: 0.5516 Iter 10089, loss: 0.5256 Iter 10090, loss: 0.4854 Iter 10091, loss: 0.5191 Iter 10092, loss: 0.5218 Iter 10093, loss: 0.4732 Iter 10094, loss: 0.5399 Iter 10095, loss: 0.5454 Iter 10096, loss: 0.4919

Iter 10097, loss: 0.4861 Iter 10098, loss: 0.5245 Iter 10099, loss: 0.5450 Iter 10100, loss: 0.5209 Iter 10101, loss: 0.5003 Iter 10102, loss: 0.5143 Iter 10103, loss: 0.4914 Iter 10104, loss: 0.5326 Iter 10105, loss: 0.5292 Iter 10106, loss: 0.5440 Iter 10107, loss: 0.5225 Iter 10108, loss: 0.5216 Iter 10109, loss: 0.5177 Iter 10110, loss: 0.5178 Iter 10111, loss: 0.5318 Iter 10112, loss: 0.5023 Iter 10113, loss: 0.5384 Iter 10114, loss: 0.5315 Iter 10115, loss: 0.5053 Iter 10116, loss: 0.5295 Iter 10117, loss: 0.5125 Iter 10118, loss: 0.5366 Iter 10119, loss: 0.5112 Iter 10120, loss: 0.5013 Iter 10121, loss: 0.5075 Iter 10122, loss: 0.5113 Iter 10123, loss: 0.5150 Iter 10124, loss: 0.5058 Iter 10125, loss: 0.5007 Iter 10126, loss: 0.4984 Iter 10127, loss: 0.5009 Iter 10128, loss: 0.5281 Iter 10129, loss: 0.5805 Iter 10130, loss: 0.5299 Iter 10131, loss: 0.5063 Iter 10132, loss: 0.5069 Iter 10133, loss: 0.4832 Iter 10134, loss: 0.5391 Iter 10135, loss: 0.5164 Iter 10136, loss: 0.5198 Iter 10137, loss: 0.4893 Iter 10138, loss: 0.5172

Iter 10139, loss: 0.5688 Iter 10140, loss: 0.5278 Iter 10141, loss: 0.4891 Iter 10142, loss: 0.5160 Iter 10143, loss: 0.4838 Iter 10144, loss: 0.5348 Iter 10145, loss: 0.5113 Iter 10146, loss: 0.5278 Iter 10147, loss: 0.5361 Iter 10148, loss: 0.5058 Iter 10149, loss: 0.5462 Iter 10150, loss: 0.4925 Iter 10151, loss: 0.4948 Iter 10152, loss: 0.4650 Iter 10153, loss: 0.4958 Iter 10154, loss: 0.5167 Iter 10155, loss: 0.5844 Iter 10156, loss: 0.5563 Iter 10157, loss: 0.5074 Iter 10158, loss: 0.5129 Iter 10159, loss: 0.5599 Iter 10160, loss: 0.5353 Iter 10161, loss: 0.4954 Iter 10162, loss: 0.5152 Iter 10163, loss: 0.5004 Iter 10164, loss: 0.5012 Iter 10165, loss: 0.5127 Iter 10166, loss: 0.5268 Iter 10167, loss: 0.5065 Iter 10168, loss: 0.5024 Iter 10169, loss: 0.5566 Iter 10170, loss: 0.4696 Iter 10171, loss: 0.5257 Iter 10172, loss: 0.5363 Iter 10173, loss: 0.5086 Iter 10174, loss: 0.5201 Iter 10175, loss: 0.5230 Iter 10176, loss: 0.4820 Iter 10177, loss: 0.5341 Iter 10178, loss: 0.5112 Iter 10179, loss: 0.5362 Iter 10180, loss: 0.5351

Iter 10181, loss: 0.5128 Iter 10182, loss: 0.5165 Iter 10183, loss: 0.5078 Iter 10184, loss: 0.4848 Iter 10185, loss: 0.5366 Iter 10186, loss: 0.5128 Iter 10187, loss: 0.4960 Iter 10188, loss: 0.5011 Iter 10189, loss: 0.5064 Iter 10190, loss: 0.4589 Iter 10191, loss: 0.5280 Iter 10192, loss: 0.4928 Iter 10193, loss: 0.4947 Iter 10194, loss: 0.5214 Iter 10195, loss: 0.5261 Iter 10196, loss: 0.5343 Iter 10197, loss: 0.5446 Iter 10198, loss: 0.5444 Iter 10199, loss: 0.5355 Iter 10200, loss: 0.5485 Iter 10201, loss: 0.5167 Iter 10202, loss: 0.5098 Iter 10203, loss: 0.4871 Iter 10204, loss: 0.4939 Iter 10205, loss: 0.5097 Iter 10206, loss: 0.4900 Iter 10207, loss: 0.5023 Iter 10208, loss: 0.5118 Iter 10209, loss: 0.5161 Iter 10210, loss: 0.5017 Iter 10211, loss: 0.5258 Iter 10212, loss: 0.5278 Iter 10213, loss: 0.4769 Iter 10214, loss: 0.5202 Iter 10215, loss: 0.5300 Iter 10216, loss: 0.4798 Iter 10217, loss: 0.5176 Iter 10218, loss: 0.4823 Iter 10219, loss: 0.5499 Iter 10220, loss: 0.5125 Iter 10221, loss: 0.5097 Iter 10222, loss: 0.5094

Iter 10223, loss: 0.5199 Iter 10224, loss: 0.5410 Iter 10225, loss: 0.5500 Iter 10226, loss: 0.5232 Iter 10227, loss: 0.5155 Iter 10228, loss: 0.5325 Iter 10229, loss: 0.4732 Iter 10230, loss: 0.4741 Iter 10231, loss: 0.5227 Iter 10232, loss: 0.4779 Iter 10233, loss: 0.5288 Iter 10234, loss: 0.5230 Iter 10235, loss: 0.5134 Iter 10236, loss: 0.4699 Iter 10237, loss: 0.5004 Iter 10238, loss: 0.5162 Iter 10239, loss: 0.5199 Iter 10240, loss: 0.5222 Iter 10241, loss: 0.4987 Iter 10242, loss: 0.5433 Iter 10243, loss: 0.5346 Iter 10244, loss: 0.5118 Iter 10245, loss: 0.5268 Iter 10246, loss: 0.5005 Iter 10247, loss: 0.5586 Iter 10248, loss: 0.5239 Iter 10249, loss: 0.4941 Iter 10250, loss: 0.4864 Iter 10251, loss: 0.5010 Iter 10252, loss: 0.5228 Iter 10253, loss: 0.4831 Iter 10254, loss: 0.4900 Iter 10255, loss: 0.4745 Iter 10256, loss: 0.5200 Iter 10257, loss: 0.5054 Iter 10258, loss: 0.4900 Iter 10259, loss: 0.5415 Iter 10260, loss: 0.5504 Iter 10261, loss: 0.4871 Iter 10262, loss: 0.5178 Iter 10263, loss: 0.5227

Iter 10264, loss: 0.5207 Iter 10265, loss: 0.5503 Iter 10266, loss: 0.5226 Iter 10267, loss: 0.5443 Iter 10268, loss: 0.5216 Iter 10269, loss: 0.5545 Iter 10270, loss: 0.5107 Iter 10271, loss: 0.5342 Iter 10272, loss: 0.4906 Iter 10273, loss: 0.5077 Iter 10274, loss: 0.5125 Iter 10275, loss: 0.5557 Iter 10276, loss: 0.5257 Iter 10277, loss: 0.5120 Iter 10278, loss: 0.4835 Iter 10279, loss: 0.5480 Iter 10280, loss: 0.5341 Iter 10281, loss: 0.5309 Iter 10282, loss: 0.4900 Iter 10283, loss: 0.4841 Iter 10284, loss: 0.5169 Iter 10285, loss: 0.5306 Iter 10286, loss: 0.5030 Iter 10287, loss: 0.5161 Iter 10288, loss: 0.5049 Iter 10289, loss: 0.4949 Iter 10290, loss: 0.5256 Iter 10291, loss: 0.5248 Iter 10292, loss: 0.4931 Iter 10293, loss: 0.5094 Iter 10294, loss: 0.4991 Iter 10295, loss: 0.5261 Iter 10296, loss: 0.5076 Iter 10297, loss: 0.5183 Iter 10298, loss: 0.5115 Iter 10299, loss: 0.5239 Iter 10300, loss: 0.4951 Iter 10301, loss: 0.5397 Iter 10302, loss: 0.5385 Iter 10303, loss: 0.5212 Iter 10304, loss: 0.5116 Iter 10305, loss: 0.5468

Iter 10306, loss: 0.4813 Iter 10307, loss: 0.5073 Iter 10308, loss: 0.4803 Iter 10309, loss: 0.5151 Iter 10310, loss: 0.5481 Iter 10311, loss: 0.5032 Iter 10312, loss: 0.5031 Iter 10313, loss: 0.5240 Iter 10314, loss: 0.5283 Iter 10315, loss: 0.4780 Iter 10316, loss: 0.5299 Iter 10317, loss: 0.4785 Iter 10318, loss: 0.5174 Iter 10319, loss: 0.5183 Iter 10320, loss: 0.5465 Iter 10321, loss: 0.5307 Iter 10322, loss: 0.5202 Iter 10323, loss: 0.5101 Iter 10324, loss: 0.4968 Iter 10325, loss: 0.4824 Iter 10326, loss: 0.4952 Iter 10327, loss: 0.5941 Iter 10328, loss: 0.5455 Iter 10329, loss: 0.5196 Iter 10330, loss: 0.5582 Iter 10331, loss: 0.5258 Iter 10332, loss: 0.5719 Iter 10333, loss: 0.5174 Iter 10334, loss: 0.5355 Iter 10335, loss: 0.5558 Iter 10336, loss: 0.4854 Iter 10337, loss: 0.5009 Iter 10338, loss: 0.5233 Iter 10339, loss: 0.5034 Iter 10340, loss: 0.5147 Iter 10341, loss: 0.5502 Iter 10342, loss: 0.5124 Iter 10343, loss: 0.4694 Iter 10344, loss: 0.5162 Iter 10345, loss: 0.5620 Iter 10346, loss: 0.5245 Iter 10347, loss: 0.4935

Iter 10348, loss: 0.4840 Iter 10349, loss: 0.5453 Iter 10350, loss: 0.5418 Iter 10351, loss: 0.4963 Iter 10352, loss: 0.5284 Iter 10353, loss: 0.5120 Iter 10354, loss: 0.5003 Iter 10355, loss: 0.5252 Iter 10356, loss: 0.4859 Iter 10357, loss: 0.5230 Iter 10358, loss: 0.5005 Iter 10359, loss: 0.5245 Iter 10360, loss: 0.5423 Iter 10361, loss: 0.5041 Iter 10362, loss: 0.4786 Iter 10363, loss: 0.5286 Iter 10364, loss: 0.5057 Iter 10365, loss: 0.5181 Iter 10366, loss: 0.5095 Iter 10367, loss: 0.5610 Iter 10368, loss: 0.5316 Iter 10369, loss: 0.5206 Iter 10370, loss: 0.5129 Iter 10371, loss: 0.4790 Iter 10372, loss: 0.4798 Iter 10373, loss: 0.5164 Iter 10374, loss: 0.5146 Iter 10375, loss: 0.4936 Iter 10376, loss: 0.5191 Iter 10377, loss: 0.5055 Iter 10378, loss: 0.5238 Iter 10379, loss: 0.5002 Iter 10380, loss: 0.5203 Iter 10381, loss: 0.4966 Iter 10382, loss: 0.5040 Iter 10383, loss: 0.5240 Iter 10384, loss: 0.5130 Iter 10385, loss: 0.4881 Iter 10386, loss: 0.5126 Iter 10387, loss: 0.5248 Iter 10388, loss: 0.5203 Iter 10389, loss: 0.5626

Iter 10390, loss: 0.4867 Iter 10391, loss: 0.5195 Iter 10392, loss: 0.5250 Iter 10393, loss: 0.5122 Iter 10394, loss: 0.5678 Iter 10395, loss: 0.4966 Iter 10396, loss: 0.5066 Iter 10397, loss: 0.4619 Iter 10398, loss: 0.5229 Iter 10399, loss: 0.5056 Iter 10400, loss: 0.5369 Iter 10401, loss: 0.4928 Iter 10402, loss: 0.5136 Iter 10403, loss: 0.5160 Iter 10404, loss: 0.4852 Iter 10405, loss: 0.5117 Iter 10406, loss: 0.4713 Iter 10407, loss: 0.4904 Iter 10408, loss: 0.5038 Iter 10409, loss: 0.5046 Iter 10410, loss: 0.4670 Iter 10411, loss: 0.5043 Iter 10412, loss: 0.5152 Iter 10413, loss: 0.5184 Iter 10414, loss: 0.4650 Iter 10415, loss: 0.5378 Iter 10416, loss: 0.5234 Iter 10417, loss: 0.5232 Iter 10418, loss: 0.5094 Iter 10419, loss: 0.5235 Iter 10420, loss: 0.5208 Iter 10421, loss: 0.4829 Iter 10422, loss: 0.4753 Iter 10423, loss: 0.5027 Iter 10424, loss: 0.5036 Iter 10425, loss: 0.5249 Iter 10426, loss: 0.4847 Iter 10427, loss: 0.5194 Iter 10428, loss: 0.4974 Iter 10429, loss: 0.4969 Iter 10430, loss: 0.5420 Iter 10431, loss: 0.5175

Iter 10432, loss: 0.5318 Iter 10433, loss: 0.5092 Iter 10434, loss: 0.4991 Iter 10435, loss: 0.5532 Iter 10436, loss: 0.5261 Iter 10437, loss: 0.5524 Iter 10438, loss: 0.5023 Iter 10439, loss: 0.5290 Iter 10440, loss: 0.5227 Iter 10441, loss: 0.5232 Iter 10442, loss: 0.4928 Iter 10443, loss: 0.5166 Iter 10444, loss: 0.5012 Iter 10445, loss: 0.5010 Iter 10446, loss: 0.5222 Iter 10447, loss: 0.5883 Iter 10448, loss: 0.4949 Iter 10449, loss: 0.5681 Iter 10450, loss: 0.5024 Iter 10451, loss: 0.4714 Iter 10452, loss: 0.4993 Iter 10453, loss: 0.5012 Iter 10454, loss: 0.5177 Iter 10455, loss: 0.5310 Iter 10456, loss: 0.4789 Iter 10457, loss: 0.5088 Iter 10458, loss: 0.4817 Iter 10459, loss: 0.5265 Iter 10460, loss: 0.5108 Iter 10461, loss: 0.5081 Iter 10462, loss: 0.5132 Iter 10463, loss: 0.5100 Iter 10464, loss: 0.5652 Iter 10465, loss: 0.4832 Iter 10466, loss: 0.4972 Iter 10467, loss: 0.4910 Iter 10468, loss: 0.5329 Iter 10469, loss: 0.5436 Iter 10470, loss: 0.5495 Iter 10471, loss: 0.5264 Iter 10472, loss: 0.4968 Iter 10473, loss: 0.5369

Iter 10474, loss: 0.5173 Iter 10475, loss: 0.5259 Iter 10476, loss: 0.5141 Iter 10477, loss: 0.4994 Iter 10478, loss: 0.5381 Iter 10479, loss: 0.5991 Iter 10480, loss: 0.4966 Iter 10481, loss: 0.5245 Iter 10482, loss: 0.5120 Iter 10483, loss: 0.5318 Iter 10484, loss: 0.4941 Iter 10485, loss: 0.5072 Iter 10486, loss: 0.5053 Iter 10487, loss: 0.4770 Iter 10488, loss: 0.4912 Iter 10489, loss: 0.5078 Iter 10490, loss: 0.5504 Iter 10491, loss: 0.5111 Iter 10492, loss: 0.5350 Iter 10493, loss: 0.5147 Iter 10494, loss: 0.5408 Iter 10495, loss: 0.4926 Iter 10496, loss: 0.5053 Iter 10497, loss: 0.4815 Iter 10498, loss: 0.5084 Iter 10499, loss: 0.5241 Iter 10500, loss: 0.5203 Iter 10501, loss: 0.4788 Iter 10502, loss: 0.5686 Iter 10503, loss: 0.4937 Iter 10504, loss: 0.5359 Iter 10505, loss: 0.5082 Iter 10506, loss: 0.4919 Iter 10507, loss: 0.4932 Iter 10508, loss: 0.5217 Iter 10509, loss: 0.5478 Iter 10510, loss: 0.5340 Iter 10511, loss: 0.5310 Iter 10512, loss: 0.5508 Iter 10513, loss: 0.5139 Iter 10514, loss: 0.4886 Iter 10515, loss: 0.5561

Iter 10516, loss: 0.4814 Iter 10517, loss: 0.4961 Iter 10518, loss: 0.4999 Iter 10519, loss: 0.5063 Iter 10520, loss: 0.5282 Iter 10521, loss: 0.4982 Iter 10522, loss: 0.5105 Iter 10523, loss: 0.5538 Iter 10524, loss: 0.5311 Iter 10525, loss: 0.5299 Iter 10526, loss: 0.5190 Iter 10527, loss: 0.5278 Iter 10528, loss: 0.5120 Iter 10529, loss: 0.5111 Iter 10530, loss: 0.5252 Iter 10531, loss: 0.4722 Iter 10532, loss: 0.5207 Iter 10533, loss: 0.4777 Iter 10534, loss: 0.4856 Iter 10535, loss: 0.5352 Iter 10536, loss: 0.4847 Iter 10537, loss: 0.5418 Iter 10538, loss: 0.5127 Iter 10539, loss: 0.4980 Iter 10540, loss: 0.5420 Iter 10541, loss: 0.4973 Iter 10542, loss: 0.4898 Iter 10543, loss: 0.5338 Iter 10544, loss: 0.5281 Iter 10545, loss: 0.5167 Iter 10546, loss: 0.4925 Iter 10547, loss: 0.5113 Iter 10548, loss: 0.5130 Iter 10549, loss: 0.5396 Iter 10550, loss: 0.5224 Iter 10551, loss: 0.5096 Iter 10552, loss: 0.5319 Iter 10553, loss: 0.5077 Iter 10554, loss: 0.5414 Iter 10555, loss: 0.5558 Iter 10556, loss: 0.4962 Iter 10557, loss: 0.4790

Iter 10558, loss: 0.5587 Iter 10559, loss: 0.5067 Iter 10560, loss: 0.5100 Iter 10561, loss: 0.5081 Iter 10562, loss: 0.5316 Iter 10563, loss: 0.4776 Iter 10564, loss: 0.5449 Iter 10565, loss: 0.5421 Iter 10566, loss: 0.5452 Iter 10567, loss: 0.5449 Iter 10568, loss: 0.5267 Iter 10569, loss: 0.5169 Iter 10570, loss: 0.4917 Iter 10571, loss: 0.5395 Iter 10572, loss: 0.4952 Iter 10573, loss: 0.5007 Iter 10574, loss: 0.5338 Iter 10575, loss: 0.5066 Iter 10576, loss: 0.5234 Iter 10577, loss: 0.5426 Iter 10578, loss: 0.5199 Iter 10579, loss: 0.5139 Iter 10580, loss: 0.5221 Iter 10581, loss: 0.5202 Iter 10582, loss: 0.5038 Iter 10583, loss: 0.4920 Iter 10584, loss: 0.5296 Iter 10585, loss: 0.4986 Iter 10586, loss: 0.5212 Iter 10587, loss: 0.4983 Iter 10588, loss: 0.5695 Iter 10589, loss: 0.4927 Iter 10590, loss: 0.4999 Iter 10591, loss: 0.5598 Iter 10592, loss: 0.5423 Iter 10593, loss: 0.5252 Iter 10594, loss: 0.4804 Iter 10595, loss: 0.4921 Iter 10596, loss: 0.5357 Iter 10597, loss: 0.5106 Iter 10598, loss: 0.5468 Iter 10599, loss: 0.5515

Iter 10600, loss: 0.5235 Iter 10601, loss: 0.5386 Iter 10602, loss: 0.5256 Iter 10603, loss: 0.4931 Iter 10604, loss: 0.5519 Iter 10605, loss: 0.4884 Iter 10606, loss: 0.5340 Iter 10607, loss: 0.5056 Iter 10608, loss: 0.4763 Iter 10609, loss: 0.5116 Iter 10610, loss: 0.4884 Iter 10611, loss: 0.5317 Iter 10612, loss: 0.5170 Iter 10613, loss: 0.5032 Iter 10614, loss: 0.5221 Iter 10615, loss: 0.5213 Iter 10616, loss: 0.4910 Iter 10617, loss: 0.5078 Iter 10618, loss: 0.5231 Iter 10619, loss: 0.5356 Iter 10620, loss: 0.4970 Iter 10621, loss: 0.5135 Iter 10622, loss: 0.5311 Iter 10623, loss: 0.5179 Iter 10624, loss: 0.5390 Iter 10625, loss: 0.5232 Iter 10626, loss: 0.5242 Iter 10627, loss: 0.4760 Iter 10628, loss: 0.4827 Iter 10629, loss: 0.5417 Iter 10630, loss: 0.5447 Iter 10631, loss: 0.5150 Iter 10632, loss: 0.5052 Iter 10633, loss: 0.4843 Iter 10634, loss: 0.5318 Iter 10635, loss: 0.4858 Iter 10636, loss: 0.5000 Iter 10637, loss: 0.5157 Iter 10638, loss: 0.4876 Iter 10639, loss: 0.5253 Iter 10640, loss: 0.5126 Iter 10641, loss: 0.5450

Iter 10642, loss: 0.5258 Iter 10643, loss: 0.4857 Iter 10644, loss: 0.5439 Iter 10645, loss: 0.5460 Iter 10646, loss: 0.5157 Iter 10647, loss: 0.5555 Iter 10648, loss: 0.4879 Iter 10649, loss: 0.5068 Iter 10650, loss: 0.4853 Iter 10651, loss: 0.4928 Iter 10652, loss: 0.5367 Iter 10653, loss: 0.5243 Iter 10654, loss: 0.4496 Iter 10655, loss: 0.5514 Iter 10656, loss: 0.4918 Iter 10657, loss: 0.4714 Iter 10658, loss: 0.4691 Iter 10659, loss: 0.5075 Iter 10660, loss: 0.4736 Iter 10661, loss: 0.4867 Iter 10662, loss: 0.4625 Iter 10663, loss: 0.5475 Iter 10664, loss: 0.5177 Iter 10665, loss: 0.4878 Iter 10666, loss: 0.4808 Iter 10667, loss: 0.5050 Iter 10668, loss: 0.5019 Iter 10669, loss: 0.4746 Iter 10670, loss: 0.5404 Iter 10671, loss: 0.5052 Iter 10672, loss: 0.4989 Iter 10673, loss: 0.4875 Iter 10674, loss: 0.5396 Iter 10675, loss: 0.4918 Iter 10676, loss: 0.4855 Iter 10677, loss: 0.5008 Iter 10678, loss: 0.5386 Iter 10679, loss: 0.5327 Iter 10680, loss: 0.4822 Iter 10681, loss: 0.4619 Iter 10682, loss: 0.5646 Iter 10683, loss: 0.4977

Iter 10684, loss: 0.4782 Iter 10685, loss: 0.4795 Iter 10686, loss: 0.5308 Iter 10687, loss: 0.5070 Iter 10688, loss: 0.5437 Iter 10689, loss: 0.4821 Iter 10690, loss: 0.4734 Iter 10691, loss: 0.4960 Iter 10692, loss: 0.5082 Iter 10693, loss: 0.5289 Iter 10694, loss: 0.5052 Iter 10695, loss: 0.5097 Iter 10696, loss: 0.5362 Iter 10697, loss: 0.4987 Iter 10698, loss: 0.5216 Iter 10699, loss: 0.5250 Iter 10700, loss: 0.4968 Iter 10701, loss: 0.5150 Iter 10702, loss: 0.4954 Iter 10703, loss: 0.4687 Iter 10704, loss: 0.5095 Iter 10705, loss: 0.4771 Iter 10706, loss: 0.5089 Iter 10707, loss: 0.4745 Iter 10708, loss: 0.5280 Iter 10709, loss: 0.5107 Iter 10710, loss: 0.4979 Iter 10711, loss: 0.5144 Iter 10712, loss: 0.4972 Iter 10713, loss: 0.5127 Iter 10714, loss: 0.5166 Iter 10715, loss: 0.5358 Iter 10716, loss: 0.5393 Iter 10717, loss: 0.5540 Iter 10718, loss: 0.5178 Iter 10719, loss: 0.5065 Iter 10720, loss: 0.5216

Iter 10721, loss: 0.5130 Iter 10722, loss: 0.4924 Iter 10723, loss: 0.5016 Iter 10724, loss: 0.5092 Iter 10725, loss: 0.5187 Iter 10726, loss: 0.5137 Iter 10727, loss: 0.4942 Iter 10728, loss: 0.5129 Iter 10729, loss: 0.5021 Iter 10730, loss: 0.5126 Iter 10731, loss: 0.4867 Iter 10732, loss: 0.5274 Iter 10733, loss: 0.5463 Iter 10734, loss: 0.4735 Iter 10735, loss: 0.5088 Iter 10736, loss: 0.5335 Iter 10737, loss: 0.5494 Iter 10738, loss: 0.4777 Iter 10739, loss: 0.4906 Iter 10740, loss: 0.5150 Iter 10741, loss: 0.4751 Iter 10742, loss: 0.4908 Iter 10743, loss: 0.5075 Iter 10744, loss: 0.4880 Iter 10745, loss: 0.5195 Iter 10746, loss: 0.4918 Iter 10747, loss: 0.5256 Iter 10748, loss: 0.4686 Iter 10749, loss: 0.5042 Iter 10750, loss: 0.4916 Iter 10751, loss: 0.5410 Iter 10752, loss: 0.4934 Iter 10753, loss: 0.5445 Iter 10754, loss: 0.5079 Iter 10755, loss: 0.4934 Iter 10756, loss: 0.5477 Iter 10757, loss: 0.5291 Iter 10758, loss: 0.4937 Iter 10759, loss: 0.5058 Iter 10760, loss: 0.5246 Iter 10761, loss: 0.5196 Iter 10762, loss: 0.4931

Iter 10763, loss: 0.5136 Iter 10764, loss: 0.5133 Iter 10765, loss: 0.4955 Iter 10766, loss: 0.5447 Iter 10767, loss: 0.5069 Iter 10768, loss: 0.4961 Iter 10769, loss: 0.5032 Iter 10770, loss: 0.5054 Iter 10771, loss: 0.5235 Iter 10772, loss: 0.5140 Iter 10773, loss: 0.5293 Iter 10774, loss: 0.5306 Iter 10775, loss: 0.5057 Iter 10776, loss: 0.4939 Iter 10777, loss: 0.5067 Iter 10778, loss: 0.5126 Iter 10779, loss: 0.5211 Iter 10780, loss: 0.4880 Iter 10781, loss: 0.5150 Iter 10782, loss: 0.4736 Iter 10783, loss: 0.4909 Iter 10784, loss: 0.5028 Iter 10785, loss: 0.5359 Iter 10786, loss: 0.4972 Iter 10787, loss: 0.5362 Iter 10788, loss: 0.5342 Iter 10789, loss: 0.5163 Iter 10790, loss: 0.5055 Iter 10791, loss: 0.4949 Iter 10792, loss: 0.5154 Iter 10793, loss: 0.4753 Iter 10794, loss: 0.5085 Iter 10795, loss: 0.5260 Iter 10796, loss: 0.5532 Iter 10797, loss: 0.4757 Iter 10798, loss: 0.5156 Iter 10799, loss: 0.5053 Iter 10800, loss: 0.5293 Iter 10801, loss: 0.4894 Iter 10802, loss: 0.5288 Iter 10803, loss: 0.4643 Iter 10804, loss: 0.5222

Iter 10805, loss: 0.5222 Iter 10806, loss: 0.5238 Iter 10807, loss: 0.5340 Iter 10808, loss: 0.4859 Iter 10809, loss: 0.5216 Iter 10810, loss: 0.4944 Iter 10811, loss: 0.5134 Iter 10812, loss: 0.5361 Iter 10813, loss: 0.5332 Iter 10814, loss: 0.5141 Iter 10815, loss: 0.4870 Iter 10816, loss: 0.5365 Iter 10817, loss: 0.4661 Iter 10818, loss: 0.5379 Iter 10819, loss: 0.5020 Iter 10820, loss: 0.4917 Iter 10821, loss: 0.4995 Iter 10822, loss: 0.5295 Iter 10823, loss: 0.5350 Iter 10824, loss: 0.5216 Iter 10825, loss: 0.5345 Iter 10826, loss: 0.5408 Iter 10827, loss: 0.5087 Iter 10828, loss: 0.4913 Iter 10829, loss: 0.5186 Iter 10830, loss: 0.4815 Iter 10831, loss: 0.4822 Iter 10832, loss: 0.4854 Iter 10833, loss: 0.4799 Iter 10834, loss: 0.4867 Iter 10835, loss: 0.5371 Iter 10836, loss: 0.4899 Iter 10837, loss: 0.5398 Iter 10838, loss: 0.4997 Iter 10839, loss: 0.4879 Iter 10840, loss: 0.5085 Iter 10841, loss: 0.5431 Iter 10842, loss: 0.5187 Iter 10843, loss: 0.5208 Iter 10844, loss: 0.5561 Iter 10845, loss: 0.5100 Iter 10846, loss: 0.5141

Iter 10847, loss: 0.4799 Iter 10848, loss: 0.4609 Iter 10849, loss: 0.5217 Iter 10850, loss: 0.5469 Iter 10851, loss: 0.4664 Iter 10852, loss: 0.4460 Iter 10853, loss: 0.5139 Iter 10854, loss: 0.5390 Iter 10855, loss: 0.5222 Iter 10856, loss: 0.5170 Iter 10857, loss: 0.5185 Iter 10858, loss: 0.5201 Iter 10859, loss: 0.5536 Iter 10860, loss: 0.5104 Iter 10861, loss: 0.5232 Iter 10862, loss: 0.4975 Iter 10863, loss: 0.4651 Iter 10864, loss: 0.5016 Iter 10865, loss: 0.5244 Iter 10866, loss: 0.4793 Iter 10867, loss: 0.5288 Iter 10868, loss: 0.4981 Iter 10869, loss: 0.5274 Iter 10870, loss: 0.5412 Iter 10871, loss: 0.4708 Iter 10872, loss: 0.4913 Iter 10873, loss: 0.5022 Iter 10874, loss: 0.5448 Iter 10875, loss: 0.5143 Iter 10876, loss: 0.5172 Iter 10877, loss: 0.5315 Iter 10878, loss: 0.5063 Iter 10879, loss: 0.5012 Iter 10880, loss: 0.5168 Iter 10881, loss: 0.5222 Iter 10882, loss: 0.4867 Iter 10883, loss: 0.5277 Iter 10884, loss: 0.4819 Iter 10885, loss: 0.5042 Iter 10886, loss: 0.5061 Iter 10887, loss: 0.5020 Iter 10888, loss: 0.5023

Iter 10889, loss: 0.4784 Iter 10890, loss: 0.5003 Iter 10891, loss: 0.4883 Iter 10892, loss: 0.5512 Iter 10893, loss: 0.5208 Iter 10894, loss: 0.4578 Iter 10895, loss: 0.4763 Iter 10896, loss: 0.5186 Iter 10897, loss: 0.5340 Iter 10898, loss: 0.5074 Iter 10899, loss: 0.4863 Iter 10900, loss: 0.5562 Iter 10901, loss: 0.5650 Iter 10902, loss: 0.5287 Iter 10903, loss: 0.5388 Iter 10904, loss: 0.5254 Iter 10905, loss: 0.5041 Iter 10906, loss: 0.5118 Iter 10907, loss: 0.5426 Iter 10908, loss: 0.5259 Iter 10909, loss: 0.4721 Iter 10910, loss: 0.4869 Iter 10911, loss: 0.4885 Iter 10912, loss: 0.4676 Iter 10913, loss: 0.4711 Iter 10914, loss: 0.4992 Iter 10915, loss: 0.5331 Iter 10916, loss: 0.5027 Iter 10917, loss: 0.4743 Iter 10918, loss: 0.5102 Iter 10919, loss: 0.4617 Iter 10920, loss: 0.4841 Iter 10921, loss: 0.5411 Iter 10922, loss: 0.5411 Iter 10923, loss: 0.5169 Iter 10924, loss: 0.5349 Iter 10925, loss: 0.4555 Iter 10926, loss: 0.4910 Iter 10927, loss: 0.5196 Iter 10928, loss: 0.5434 Iter 10929, loss: 0.5527 Iter 10930, loss: 0.4943

Iter 10931, loss: 0.5142 Iter 10932, loss: 0.5300 Iter 10933, loss: 0.4779 Iter 10934, loss: 0.5100 Iter 10935, loss: 0.5252 Iter 10936, loss: 0.5512 Iter 10937, loss: 0.5154 Iter 10938, loss: 0.5196 Iter 10939, loss: 0.4990 Iter 10940, loss: 0.5027 Iter 10941, loss: 0.5274 Iter 10942, loss: 0.4894 Iter 10943, loss: 0.4696 Iter 10944, loss: 0.5337 Iter 10945, loss: 0.4607 Iter 10946, loss: 0.5356 Iter 10947, loss: 0.4974 Iter 10948, loss: 0.4873 Iter 10949, loss: 0.5353 Iter 10950, loss: 0.4919 Iter 10951, loss: 0.5066 Iter 10952, loss: 0.5088 Iter 10953, loss: 0.4668 Iter 10954, loss: 0.5100 Iter 10955, loss: 0.5344 Iter 10956, loss: 0.5340 Iter 10957, loss: 0.5235 Iter 10958, loss: 0.4844 Iter 10959, loss: 0.5033 Iter 10960, loss: 0.5273 Iter 10961, loss: 0.5004 Iter 10962, loss: 0.4616 Iter 10963, loss: 0.4993 Iter 10964, loss: 0.4992 Iter 10965, loss: 0.5062 Iter 10966, loss: 0.5234 Iter 10967, loss: 0.5160 Iter 10968, loss: 0.4985 Iter 10969, loss: 0.5215 Iter 10970, loss: 0.4904 Iter 10971, loss: 0.5264 Iter 10972, loss: 0.5022

Iter 10973, loss: 0.4547 Iter 10974, loss: 0.5310 Iter 10975, loss: 0.4969 Iter 10976, loss: 0.4806 Iter 10977, loss: 0.4598 Iter 10978, loss: 0.5001 Iter 10979, loss: 0.5167 Iter 10980, loss: 0.4945 Iter 10981, loss: 0.4858 Iter 10982, loss: 0.5444 Iter 10983, loss: 0.4628 Iter 10984, loss: 0.5145 Iter 10985, loss: 0.5240 Iter 10986, loss: 0.5312 Iter 10987, loss: 0.4967 Iter 10988, loss: 0.5015 Iter 10989, loss: 0.5344 Iter 10990, loss: 0.4720 Iter 10991, loss: 0.5237 Iter 10992, loss: 0.4975 Iter 10993, loss: 0.5209 Iter 10994, loss: 0.5487 Iter 10995, loss: 0.4802 Iter 10996, loss: 0.4741 Iter 10997, loss: 0.4857 Iter 10998, loss: 0.5113 Iter 10999, loss: 0.5461 Iter 11000, loss: 0.5398 Iter 11001, loss: 0.4855 Iter 11002, loss: 0.5239 Iter 11003, loss: 0.5134 Iter 11004, loss: 0.4757 Iter 11005, loss: 0.4772 Iter 11006, loss: 0.5109 Iter 11007, loss: 0.5209 Iter 11008, loss: 0.5120 Iter 11009, loss: 0.5194 Iter 11010, loss: 0.4720 Iter 11011, loss: 0.5001 Iter 11012, loss: 0.4932 Iter 11013, loss: 0.5049 Iter 11014, loss: 0.5077

Iter 11015, loss: 0.5323 Iter 11016, loss: 0.5109 Iter 11017, loss: 0.5380 Iter 11018, loss: 0.5064 Iter 11019, loss: 0.5444 Iter 11020, loss: 0.5178 Iter 11021, loss: 0.5020 Iter 11022, loss: 0.5369 Iter 11023, loss: 0.4964 Iter 11024, loss: 0.5176 Iter 11025, loss: 0.4396 Iter 11026, loss: 0.5218 Iter 11027, loss: 0.4978 Iter 11028, loss: 0.5070 Iter 11029, loss: 0.4833 Iter 11030, loss: 0.5025 Iter 11031, loss: 0.5147 Iter 11032, loss: 0.5204 Iter 11033, loss: 0.4713 Iter 11034, loss: 0.5144 Iter 11035, loss: 0.4554 Iter 11036, loss: 0.5113 Iter 11037, loss: 0.5065 Iter 11038, loss: 0.5076 Iter 11039, loss: 0.5225 Iter 11040, loss: 0.5021 Iter 11041, loss: 0.5078 Iter 11042, loss: 0.5430 Iter 11043, loss: 0.5220 Iter 11044, loss: 0.5363 Iter 11045, loss: 0.4718 Iter 11046, loss: 0.5187 Iter 11047, loss: 0.5053 Iter 11048, loss: 0.5119 Iter 11049, loss: 0.4641 Iter 11050, loss: 0.4820 Iter 11051, loss: 0.5030 Iter 11052, loss: 0.5098 Iter 11053, loss: 0.4405 Iter 11054, loss: 0.4813 Iter 11055, loss: 0.5421 Iter 11056, loss: 0.4976

Iter 11057, loss: 0.5231 Iter 11058, loss: 0.5056 Iter 11059, loss: 0.5198 Iter 11060, loss: 0.4715 Iter 11061, loss: 0.5253 Iter 11062, loss: 0.4903 Iter 11063, loss: 0.5172 Iter 11064, loss: 0.5020 Iter 11065, loss: 0.5098 Iter 11066, loss: 0.4877 Iter 11067, loss: 0.4894 Iter 11068, loss: 0.5150 Iter 11069, loss: 0.5082 Iter 11070, loss: 0.5311 Iter 11071, loss: 0.4830 Iter 11072, loss: 0.5856 Iter 11073, loss: 0.5331 Iter 11074, loss: 0.5060 Iter 11075, loss: 0.5059 Iter 11076, loss: 0.4846 Iter 11077, loss: 0.4943 Iter 11078, loss: 0.5626 Iter 11079, loss: 0.4981 Iter 11080, loss: 0.5561 Iter 11081, loss: 0.5292 Iter 11082, loss: 0.5354 Iter 11083, loss: 0.5369 Iter 11084, loss: 0.5088 Iter 11085, loss: 0.5245 Iter 11086, loss: 0.5270 Iter 11087, loss: 0.5180 Iter 11088, loss: 0.5100 Iter 11089, loss: 0.5083 Iter 11090, loss: 0.5130 Iter 11091, loss: 0.5631 Iter 11092, loss: 0.5201 Iter 11093, loss: 0.5206 Iter 11094, loss: 0.4958 Iter 11095, loss: 0.5213 Iter 11096, loss: 0.4877 Iter 11097, loss: 0.5655 Iter 11098, loss: 0.4686

Iter 11099, loss: 0.4961 Iter 11100, loss: 0.5191 Iter 11101, loss: 0.4677 Iter 11102, loss: 0.5201 Iter 11103, loss: 0.4957 Iter 11104, loss: 0.4982 Iter 11105, loss: 0.5107 Iter 11106, loss: 0.4919 Iter 11107, loss: 0.4824 Iter 11108, loss: 0.4799 Iter 11109, loss: 0.5066 Iter 11110, loss: 0.4882 Iter 11111, loss: 0.5147 Iter 11112, loss: 0.4957 Iter 11113, loss: 0.5187 Iter 11114, loss: 0.5495 Iter 11115, loss: 0.5315 Iter 11116, loss: 0.4668 Iter 11117, loss: 0.4580 Iter 11118, loss: 0.5151 Iter 11119, loss: 0.4976 Iter 11120, loss: 0.5271 Iter 11121, loss: 0.5147 Iter 11122, loss: 0.5091 Iter 11123, loss: 0.4848 Iter 11124, loss: 0.4823 Iter 11125, loss: 0.5214 Iter 11126, loss: 0.4909 Iter 11127, loss: 0.4844 Iter 11128, loss: 0.4936 Iter 11129, loss: 0.5032 Iter 11130, loss: 0.5189 Iter 11131, loss: 0.4998 Iter 11132, loss: 0.4880 Iter 11133, loss: 0.4887 Iter 11134, loss: 0.5136 Iter 11135, loss: 0.5375 Iter 11136, loss: 0.5526 Iter 11137, loss: 0.4991 Iter 11138, loss: 0.5167 Iter 11139, loss: 0.5036 Iter 11140, loss: 0.5234

Iter 11141, loss: 0.5141 Iter 11142, loss: 0.4785 Iter 11143, loss: 0.5042 Iter 11144, loss: 0.5460 Iter 11145, loss: 0.4976 Iter 11146, loss: 0.5241 Iter 11147, loss: 0.4980 Iter 11148, loss: 0.4717 Iter 11149, loss: 0.4895 Iter 11150, loss: 0.4978 Iter 11151, loss: 0.5378 Iter 11152, loss: 0.5065 Iter 11153, loss: 0.5166 Iter 11154, loss: 0.5371 Iter 11155, loss: 0.5192 Iter 11156, loss: 0.5122 Iter 11157, loss: 0.5261 Iter 11158, loss: 0.5231 Iter 11159, loss: 0.5087 Iter 11160, loss: 0.4968 Iter 11161, loss: 0.5000 Iter 11162, loss: 0.5401 Iter 11163, loss: 0.4942 Iter 11164, loss: 0.5030 Iter 11165, loss: 0.5265 Iter 11166, loss: 0.4859 Iter 11167, loss: 0.5038 Iter 11168, loss: 0.4860 Iter 11169, loss: 0.4961 Iter 11170, loss: 0.5153 Iter 11171, loss: 0.5349 Iter 11172, loss: 0.5194 Iter 11173, loss: 0.4906 Iter 11174, loss: 0.5009 Iter 11175, loss: 0.5172 Iter 11176, loss: 0.4943 Iter 11177, loss: 0.4901 Iter 11178, loss: 0.5423 Iter 11179, loss: 0.5165 Iter 11180, loss: 0.5030 Iter 11181, loss: 0.5048 Iter 11182, loss: 0.4843

Iter 11183, loss: 0.4919 Iter 11184, loss: 0.5045 Iter 11185, loss: 0.5209 Iter 11186, loss: 0.5086 Iter 11187, loss: 0.5005 Iter 11188, loss: 0.4819 Iter 11189, loss: 0.4681 Iter 11190, loss: 0.5114 Iter 11191, loss: 0.5279 Iter 11192, loss: 0.5227 Iter 11193, loss: 0.5238 Iter 11194, loss: 0.4907 Iter 11195, loss: 0.5204 Iter 11196, loss: 0.4653 Iter 11197, loss: 0.5037 Iter 11198, loss: 0.4951 Iter 11199, loss: 0.5569 Iter 11200, loss: 0.5192 Iter 11201, loss: 0.4907 Iter 11202, loss: 0.4999 Iter 11203, loss: 0.5380 Iter 11204, loss: 0.5376 Iter 11205, loss: 0.4941 Iter 11206, loss: 0.4848 Iter 11207, loss: 0.5246 Iter 11208, loss: 0.5042 Iter 11209, loss: 0.4520 Iter 11210, loss: 0.4972 Iter 11211, loss: 0.5013 Iter 11212, loss: 0.4848 Iter 11213, loss: 0.4913 Iter 11214, loss: 0.5085 Iter 11215, loss: 0.4920 Iter 11216, loss: 0.4998 Iter 11217, loss: 0.5014 Iter 11218, loss: 0.5071 Iter 11219, loss: 0.5126 Iter 11220, loss: 0.5180 Iter 11221, loss: 0.5164 Iter 11222, loss: 0.5246 Iter 11223, loss: 0.5277 Iter 11224, loss: 0.5218

Iter 11225, loss: 0.4997 Iter 11226, loss: 0.5163 Iter 11227, loss: 0.5375 Iter 11228, loss: 0.5500 Iter 11229, loss: 0.4758 Iter 11230, loss: 0.4480 Iter 11231, loss: 0.5133 Iter 11232, loss: 0.5231 Iter 11233, loss: 0.5291 Iter 11234, loss: 0.5120 Iter 11235, loss: 0.5065 Iter 11236, loss: 0.4753 Iter 11237, loss: 0.4835 Iter 11238, loss: 0.4887 Iter 11239, loss: 0.4797 Iter 11240, loss: 0.5071 Iter 11241, loss: 0.5022 Iter 11242, loss: 0.4987 Iter 11243, loss: 0.4826 Iter 11244, loss: 0.4882 Iter 11245, loss: 0.5085 Iter 11246, loss: 0.5114 Iter 11247, loss: 0.4490 Iter 11248, loss: 0.5247 Iter 11249, loss: 0.5204 Iter 11250, loss: 0.5482 Iter 11251, loss: 0.5257 Iter 11252, loss: 0.5369 Iter 11253, loss: 0.5220 Iter 11254, loss: 0.5068 Iter 11255, loss: 0.5246 Iter 11256, loss: 0.4691 Iter 11257, loss: 0.4790 Iter 11258, loss: 0.4866 Iter 11259, loss: 0.4880 Iter 11260, loss: 0.5376 Iter 11261, loss: 0.4886 Iter 11262, loss: 0.4992 Iter 11263, loss: 0.5091 Iter 11264, loss: 0.5012 Iter 11265, loss: 0.4888 Iter 11266, loss: 0.4945

Iter 11267, loss: 0.4790 Iter 11268, loss: 0.5140 Iter 11269, loss: 0.4757 Iter 11270, loss: 0.4953 Iter 11271, loss: 0.5152 Iter 11272, loss: 0.5248 Iter 11273, loss: 0.4832 Iter 11274, loss: 0.5031 Iter 11275, loss: 0.5110 Iter 11276, loss: 0.5145 Iter 11277, loss: 0.5183 Iter 11278, loss: 0.5000 Iter 11279, loss: 0.4988 Iter 11280, loss: 0.5265 Iter 11281, loss: 0.5390 Iter 11282, loss: 0.5084 Iter 11283, loss: 0.5141 Iter 11284, loss: 0.5221 Iter 11285, loss: 0.4924 Iter 11286, loss: 0.5210 Iter 11287, loss: 0.4886 Iter 11288, loss: 0.4895 Iter 11289, loss: 0.5007 Iter 11290, loss: 0.4608 Iter 11291, loss: 0.4878 Iter 11292, loss: 0.4864 Iter 11293, loss: 0.5353 Iter 11294, loss: 0.5076 Iter 11295, loss: 0.5049 Iter 11296, loss: 0.5270 Iter 11297, loss: 0.4802 Iter 11298, loss: 0.4933 Iter 11299, loss: 0.5126 Iter 11300, loss: 0.4883 Iter 11301, loss: 0.5419 Iter 11302, loss: 0.4952 Iter 11303, loss: 0.5247 Iter 11304, loss: 0.4779 Iter 11305, loss: 0.5148 Iter 11306, loss: 0.4711 Iter 11307, loss: 0.5415 Iter 11308, loss: 0.5257

Iter 11309, loss: 0.5394 Iter 11310, loss: 0.4965 Iter 11311, loss: 0.4811 Iter 11312, loss: 0.4761 Iter 11313, loss: 0.4700 Iter 11314, loss: 0.5092 Iter 11315, loss: 0.5068 Iter 11316, loss: 0.4953 Iter 11317, loss: 0.5089 Iter 11318, loss: 0.5128 Iter 11319, loss: 0.5272 Iter 11320, loss: 0.5078 Iter 11321, loss: 0.4938 Iter 11322, loss: 0.5164 Iter 11323, loss: 0.5138 Iter 11324, loss: 0.5136 Iter 11325, loss: 0.5384 Iter 11326, loss: 0.5235 Iter 11327, loss: 0.4963 Iter 11328, loss: 0.4999 Iter 11329, loss: 0.5086 Iter 11330, loss: 0.5024 Iter 11331, loss: 0.4957 Iter 11332, loss: 0.5447 Iter 11333, loss: 0.4771 Iter 11334, loss: 0.5090 Iter 11335, loss: 0.4968 Iter 11336, loss: 0.4596 Iter 11337, loss: 0.5295 Iter 11338, loss: 0.5185 Iter 11339, loss: 0.4648 Iter 11340, loss: 0.4953 Iter 11341, loss: 0.4965 Iter 11342, loss: 0.5341 Iter 11343, loss: 0.5230 Iter 11344, loss: 0.4788 Iter 11345, loss: 0.4769 Iter 11346, loss: 0.4969 Iter 11347, loss: 0.5593 Iter 11348, loss: 0.4486 Iter 11349, loss: 0.4638 Iter 11350, loss: 0.5202

Iter 11351, loss: 0.5134 Iter 11352, loss: 0.4889 Iter 11353, loss: 0.5294 Iter 11354, loss: 0.5408 Iter 11355, loss: 0.4720 Iter 11356, loss: 0.5104 Iter 11357, loss: 0.4885 Iter 11358, loss: 0.4855 Iter 11359, loss: 0.4732 Iter 11360, loss: 0.5259 Iter 11361, loss: 0.5007 Iter 11362, loss: 0.5117 Iter 11363, loss: 0.5021 Iter 11364, loss: 0.5075 Iter 11365, loss: 0.4804 Iter 11366, loss: 0.4796 Iter 11367, loss: 0.4960 Iter 11368, loss: 0.5005 Iter 11369, loss: 0.5566 Iter 11370, loss: 0.4864 Iter 11371, loss: 0.4909 Iter 11372, loss: 0.4977 Iter 11373, loss: 0.4997 Iter 11374, loss: 0.4876 Iter 11375, loss: 0.5248 Iter 11376, loss: 0.5482 Iter 11377, loss: 0.4744 Iter 11378, loss: 0.5293 Iter 11379, loss: 0.4502 Iter 11380, loss: 0.4680 Iter 11381, loss: 0.4783 Iter 11382, loss: 0.5291 Iter 11383, loss: 0.5198 Iter 11384, loss: 0.5221 Iter 11385, loss: 0.5086 Iter 11386, loss: 0.5197 Iter 11387, loss: 0.4826 Iter 11388, loss: 0.5161 Iter 11389, loss: 0.4882 Iter 11390, loss: 0.5121 Iter 11391, loss: 0.4864 Iter 11392, loss: 0.5227

Iter 11393, loss: 0.5034 Iter 11394, loss: 0.5238 Iter 11395, loss: 0.4856 Iter 11396, loss: 0.5185 Iter 11397, loss: 0.5287 Iter 11398, loss: 0.5430 Iter 11399, loss: 0.4909 Iter 11400, loss: 0.4746 Iter 11401, loss: 0.4890 Iter 11402, loss: 0.5052 Iter 11403, loss: 0.4751 Iter 11404, loss: 0.5196 Iter 11405, loss: 0.5394 Iter 11406, loss: 0.4926 Iter 11407, loss: 0.5024 Iter 11408, loss: 0.4840 Iter 11409, loss: 0.5141 Iter 11410, loss: 0.4545 Iter 11411, loss: 0.4994 Iter 11412, loss: 0.4876 Iter 11413, loss: 0.4821 Iter 11414, loss: 0.5364 Iter 11415, loss: 0.5120 Iter 11416, loss: 0.5134 Iter 11417, loss: 0.4615 Iter 11418, loss: 0.5083 Iter 11419, loss: 0.5214 Iter 11420, loss: 0.5018 Iter 11421, loss: 0.5261 Iter 11422, loss: 0.4992 Iter 11423, loss: 0.5039 Iter 11424, loss: 0.5104 Iter 11425, loss: 0.5111 Iter 11426, loss: 0.5226 Iter 11427, loss: 0.4949 Iter 11428, loss: 0.4800 Iter 11429, loss: 0.5246 Iter 11430, loss: 0.4939 Iter 11431, loss: 0.5389 Iter 11432, loss: 0.5303 Iter 11433, loss: 0.5145 Iter 11434, loss: 0.4847

Iter 11435, loss: 0.4896 Iter 11436, loss: 0.4859 Iter 11437, loss: 0.4620 Iter 11438, loss: 0.5079 Iter 11439, loss: 0.5096 Iter 11440, loss: 0.4800 Iter 11441, loss: 0.4842 Iter 11442, loss: 0.5159 Iter 11443, loss: 0.4509 Iter 11444, loss: 0.5596 Iter 11445, loss: 0.5049 Iter 11446, loss: 0.4668 Iter 11447, loss: 0.5164 Iter 11448, loss: 0.5223 Iter 11449, loss: 0.5133 Iter 11450, loss: 0.5101 Iter 11451, loss: 0.4985 Iter 11452, loss: 0.4875 Iter 11453, loss: 0.5036 Iter 11454, loss: 0.5305 Iter 11455, loss: 0.5162 Iter 11456, loss: 0.5412 Iter 11457, loss: 0.5337 Iter 11458, loss: 0.4647 Iter 11459, loss: 0.5211 Iter 11460, loss: 0.4816 Iter 11461, loss: 0.5009 Iter 11462, loss: 0.5152 Iter 11463, loss: 0.5235 Iter 11464, loss: 0.5004 Iter 11465, loss: 0.4857 Iter 11466, loss: 0.5195 Iter 11467, loss: 0.5054 Iter 11468, loss: 0.4951 Iter 11469, loss: 0.4981 Iter 11470, loss: 0.5400 Iter 11471, loss: 0.4809 Iter 11472, loss: 0.4819 Iter 11473, loss: 0.4892 Iter 11474, loss: 0.5104 Iter 11475, loss: 0.5249 Iter 11476, loss: 0.4663

Iter 11477, loss: 0.5160 Iter 11478, loss: 0.4533 Iter 11479, loss: 0.4981 Iter 11480, loss: 0.5235 Iter 11481, loss: 0.4960 Iter 11482, loss: 0.5198 Iter 11483, loss: 0.5084 Iter 11484, loss: 0.5569 Iter 11485, loss: 0.4673 Iter 11486, loss: 0.5132 Iter 11487, loss: 0.5355 Iter 11488, loss: 0.5306 Iter 11489, loss: 0.4967 Iter 11490, loss: 0.5052 Iter 11491, loss: 0.4942 Iter 11492, loss: 0.5180 Iter 11493, loss: 0.5027 Iter 11494, loss: 0.5296 Iter 11495, loss: 0.5110 Iter 11496, loss: 0.5405 Iter 11497, loss: 0.5215 Iter 11498, loss: 0.4672 Iter 11499, loss: 0.5413 Iter 11500, loss: 0.4740 Iter 11501, loss: 0.5031 Iter 11502, loss: 0.4910 Iter 11503, loss: 0.4812 Iter 11504, loss: 0.5294 Iter 11505, loss: 0.5119 Iter 11506, loss: 0.4713 Iter 11507, loss: 0.5080 Iter 11508, loss: 0.5057 Iter 11509, loss: 0.5109 Iter 11510, loss: 0.4960 Iter 11511, loss: 0.4812 Iter 11512, loss: 0.5296 Iter 11513, loss: 0.4915 Iter 11514, loss: 0.4989 Iter 11515, loss: 0.4915 Iter 11516, loss: 0.5039

Iter 11517, loss: 0.5097 Iter 11518, loss: 0.5008 Iter 11519, loss: 0.5527 Iter 11520, loss: 0.5540 Iter 11521, loss: 0.5494 Iter 11522, loss: 0.4675 Iter 11523, loss: 0.4911 Iter 11524, loss: 0.5100 Iter 11525, loss: 0.4845 Iter 11526, loss: 0.5228 Iter 11527, loss: 0.5060 Iter 11528, loss: 0.4922 Iter 11529, loss: 0.4849 Iter 11530, loss: 0.4904 Iter 11531, loss: 0.5524 Iter 11532, loss: 0.5086 Iter 11533, loss: 0.5225 Iter 11534, loss: 0.5016 Iter 11535, loss: 0.4726 Iter 11536, loss: 0.5104 Iter 11537, loss: 0.4890 Iter 11538, loss: 0.4803 Iter 11539, loss: 0.5097 Iter 11540, loss: 0.4974 Iter 11541, loss: 0.4877 Iter 11542, loss: 0.4889 Iter 11543, loss: 0.5059 Iter 11544, loss: 0.5479 Iter 11545, loss: 0.5160 Iter 11546, loss: 0.5739 Iter 11547, loss: 0.5146 Iter 11548, loss: 0.5013 Iter 11549, loss: 0.5290 Iter 11550, loss: 0.5278 Iter 11551, loss: 0.5076 Iter 11552, loss: 0.5339 Iter 11553, loss: 0.4729 Iter 11554, loss: 0.5229 Iter 11555, loss: 0.4614 Iter 11556, loss: 0.4760 Iter 11557, loss: 0.5236

Iter 11558, loss: 0.5305 Iter 11559, loss: 0.5013 Iter 11560, loss: 0.5390 Iter 11561, loss: 0.4908 Iter 11562, loss: 0.4794 Iter 11563, loss: 0.5100 Iter 11564, loss: 0.5258 Iter 11565, loss: 0.5137 Iter 11566, loss: 0.5059 Iter 11567, loss: 0.4877 Iter 11568, loss: 0.4900 Iter 11569, loss: 0.4772 Iter 11570, loss: 0.4681 Iter 11571, loss: 0.5085 Iter 11572, loss: 0.4828 Iter 11573, loss: 0.5255 Iter 11574, loss: 0.4705 Iter 11575, loss: 0.4928 Iter 11576, loss: 0.4858 Iter 11577, loss: 0.4990 Iter 11578, loss: 0.4857 Iter 11579, loss: 0.5133 Iter 11580, loss: 0.4779 Iter 11581, loss: 0.5069 Iter 11582, loss: 0.5230 Iter 11583, loss: 0.5157 Iter 11584, loss: 0.5429 Iter 11585, loss: 0.4969 Iter 11586, loss: 0.5165 Iter 11587, loss: 0.5228 Iter 11588, loss: 0.5368 Iter 11589, loss: 0.4804 Iter 11590, loss: 0.5073 Iter 11591, loss: 0.4855 Iter 11592, loss: 0.5302 Iter 11593, loss: 0.4792 Iter 11594, loss: 0.5000 Iter 11595, loss: 0.5061 Iter 11596, loss: 0.5152 Iter 11597, loss: 0.5372 Iter 11598, loss: 0.4749 Iter 11599, loss: 0.5008

Iter 11600, loss: 0.5125 Iter 11601, loss: 0.5190 Iter 11602, loss: 0.4901 Iter 11603, loss: 0.4670 Iter 11604, loss: 0.4659 Iter 11605, loss: 0.5359 Iter 11606, loss: 0.5280 Iter 11607, loss: 0.5049 Iter 11608, loss: 0.5149 Iter 11609, loss: 0.4775 Iter 11610, loss: 0.4838 Iter 11611, loss: 0.5040 Iter 11612, loss: 0.5027 Iter 11613, loss: 0.5168 Iter 11614, loss: 0.5300 Iter 11615, loss: 0.4808 Iter 11616, loss: 0.4687 Iter 11617, loss: 0.5218 Iter 11618, loss: 0.5323 Iter 11619, loss: 0.5351 Iter 11620, loss: 0.4918 Iter 11621, loss: 0.4771 Iter 11622, loss: 0.5297 Iter 11623, loss: 0.4813 Iter 11624, loss: 0.5437 Iter 11625, loss: 0.5127 Iter 11626, loss: 0.4995 Iter 11627, loss: 0.5253 Iter 11628, loss: 0.4874 Iter 11629, loss: 0.4927 Iter 11630, loss: 0.5133 Iter 11631, loss: 0.4818 Iter 11632, loss: 0.4827 Iter 11633, loss: 0.4966 Iter 11634, loss: 0.4786 Iter 11635, loss: 0.4805 Iter 11636, loss: 0.5433 Iter 11637, loss: 0.5004 Iter 11638, loss: 0.5006 Iter 11639, loss: 0.4847 Iter 11640, loss: 0.4966 Iter 11641, loss: 0.4689

Iter 11642, loss: 0.4825 Iter 11643, loss: 0.4867 Iter 11644, loss: 0.5222 Iter 11645, loss: 0.5410 Iter 11646, loss: 0.4657 Iter 11647, loss: 0.5061 Iter 11648, loss: 0.4953 Iter 11649, loss: 0.5122 Iter 11650, loss: 0.5149 Iter 11651, loss: 0.4686 Iter 11652, loss: 0.4996 Iter 11653, loss: 0.5049 Iter 11654, loss: 0.5338 Iter 11655, loss: 0.5192 Iter 11656, loss: 0.5236 Iter 11657, loss: 0.4975 Iter 11658, loss: 0.4980 Iter 11659, loss: 0.5313 Iter 11660, loss: 0.5113 Iter 11661, loss: 0.5287 Iter 11662, loss: 0.4715 Iter 11663, loss: 0.4885 Iter 11664, loss: 0.5352 Iter 11665, loss: 0.5067 Iter 11666, loss: 0.5273 Iter 11667, loss: 0.5058 Iter 11668, loss: 0.5089 Iter 11669, loss: 0.5369 Iter 11670, loss: 0.5267 Iter 11671, loss: 0.5103 Iter 11672, loss: 0.4732 Iter 11673, loss: 0.4938 Iter 11674, loss: 0.5366 Iter 11675, loss: 0.5225 Iter 11676, loss: 0.5291 Iter 11677, loss: 0.5357 Iter 11678, loss: 0.5147 Iter 11679, loss: 0.5280 Iter 11680, loss: 0.4915 Iter 11681, loss: 0.5022 Iter 11682, loss: 0.4841 Iter 11683, loss: 0.5468

Iter 11684, loss: 0.4939 Iter 11685, loss: 0.5371 Iter 11686, loss: 0.5167 Iter 11687, loss: 0.5214 Iter 11688, loss: 0.4818 Iter 11689, loss: 0.4812 Iter 11690, loss: 0.5051 Iter 11691, loss: 0.5185 Iter 11692, loss: 0.4648 Iter 11693, loss: 0.5124 Iter 11694, loss: 0.4815 Iter 11695, loss: 0.5049 Iter 11696, loss: 0.5022 Iter 11697, loss: 0.4837 Iter 11698, loss: 0.4850 Iter 11699, loss: 0.5139 Iter 11700, loss: 0.5423 Iter 11701, loss: 0.5394 Iter 11702, loss: 0.4995 Iter 11703, loss: 0.4942 Iter 11704, loss: 0.4722 Iter 11705, loss: 0.4900 Iter 11706, loss: 0.4872 Iter 11707, loss: 0.5008 Iter 11708, loss: 0.5190 Iter 11709, loss: 0.5204 Iter 11710, loss: 0.4775 Iter 11711, loss: 0.5031 Iter 11712, loss: 0.5194 Iter 11713, loss: 0.4836 Iter 11714, loss: 0.4749 Iter 11715, loss: 0.4947 Iter 11716, loss: 0.5220 Iter 11717, loss: 0.5124 Iter 11718, loss: 0.5290 Iter 11719, loss: 0.5160 Iter 11720, loss: 0.5349 Iter 11721, loss: 0.4718 Iter 11722, loss: 0.5417 Iter 11723, loss: 0.5118 Iter 11724, loss: 0.4834 Iter 11725, loss: 0.4964

Iter 11726, loss: 0.5326 Iter 11727, loss: 0.4720 Iter 11728, loss: 0.5802 Iter 11729, loss: 0.4930 Iter 11730, loss: 0.5072 Iter 11731, loss: 0.4859 Iter 11732, loss: 0.5267 Iter 11733, loss: 0.4969 Iter 11734, loss: 0.4904 Iter 11735, loss: 0.4818 Iter 11736, loss: 0.5096 Iter 11737, loss: 0.5083 Iter 11738, loss: 0.4666 Iter 11739, loss: 0.4648 Iter 11740, loss: 0.4585 Iter 11741, loss: 0.5172 Iter 11742, loss: 0.5354 Iter 11743, loss: 0.4668 Iter 11744, loss: 0.5151 Iter 11745, loss: 0.5233 Iter 11746, loss: 0.4839 Iter 11747, loss: 0.4561 Iter 11748, loss: 0.5078 Iter 11749, loss: 0.5458 Iter 11750, loss: 0.4957 Iter 11751, loss: 0.4778 Iter 11752, loss: 0.5061 Iter 11753, loss: 0.4752 Iter 11754, loss: 0.4466 Iter 11755, loss: 0.4979 Iter 11756, loss: 0.5353 Iter 11757, loss: 0.5528 Iter 11758, loss: 0.4960 Iter 11759, loss: 0.4990 Iter 11760, loss: 0.5233 Iter 11761, loss: 0.4726 Iter 11762, loss: 0.4708 Iter 11763, loss: 0.5092 Iter 11764, loss: 0.5501 Iter 11765, loss: 0.4736 Iter 11766, loss: 0.4801

Iter 11767, loss: 0.5167 Iter 11768, loss: 0.4711 Iter 11769, loss: 0.4842 Iter 11770, loss: 0.5169 Iter 11771, loss: 0.5049 Iter 11772, loss: 0.5025 Iter 11773, loss: 0.5383 Iter 11774, loss: 0.5345 Iter 11775, loss: 0.5061 Iter 11776, loss: 0.5250 Iter 11777, loss: 0.4530 Iter 11778, loss: 0.5211 Iter 11779, loss: 0.4896 Iter 11780, loss: 0.5212 Iter 11781, loss: 0.5057 Iter 11782, loss: 0.5053 Iter 11783, loss: 0.4609 Iter 11784, loss: 0.5170 Iter 11785, loss: 0.5198 Iter 11786, loss: 0.5407 Iter 11787, loss: 0.4759 Iter 11788, loss: 0.5091 Iter 11789, loss: 0.4868 Iter 11790, loss: 0.4608 Iter 11791, loss: 0.5026 Iter 11792, loss: 0.4777 Iter 11793, loss: 0.5029 Iter 11794, loss: 0.5547 Iter 11795, loss: 0.5388 Iter 11796, loss: 0.5106 Iter 11797, loss: 0.4830 Iter 11798, loss: 0.4999 Iter 11799, loss: 0.5207 Iter 11800, loss: 0.5014 Iter 11801, loss: 0.4749 Iter 11802, loss: 0.5141 Iter 11803, loss: 0.5340 Iter 11804, loss: 0.4691 Iter 11805, loss: 0.5322 Iter 11806, loss: 0.4664 Iter 11807, loss: 0.5046

Iter 11808, loss: 0.4781 Iter 11809, loss: 0.5266 Iter 11810, loss: 0.4843 Iter 11811, loss: 0.4976 Iter 11812, loss: 0.5186 Iter 11813, loss: 0.4967 Iter 11814, loss: 0.5179 Iter 11815, loss: 0.5381 Iter 11816, loss: 0.4860 Iter 11817, loss: 0.4677 Iter 11818, loss: 0.5275 Iter 11819, loss: 0.4720 Iter 11820, loss: 0.4731 Iter 11821, loss: 0.5005 Iter 11822, loss: 0.5570 Iter 11823, loss: 0.5047 Iter 11824, loss: 0.4927 Iter 11825, loss: 0.5143 Iter 11826, loss: 0.4752 Iter 11827, loss: 0.4985 Iter 11828, loss: 0.5416 Iter 11829, loss: 0.5163 Iter 11830, loss: 0.5080 Iter 11831, loss: 0.5486 Iter 11832, loss: 0.4729 Iter 11833, loss: 0.5193 Iter 11834, loss: 0.5212 Iter 11835, loss: 0.5421 Iter 11836, loss: 0.4516 Iter 11837, loss: 0.5143 Iter 11838, loss: 0.5078 Iter 11839, loss: 0.4847 Iter 11840, loss: 0.5044 Iter 11841, loss: 0.4548 Iter 11842, loss: 0.5196 Iter 11843, loss: 0.5184 Iter 11844, loss: 0.4855 Iter 11845, loss: 0.5471 Iter 11846, loss: 0.4827 Iter 11847, loss: 0.5115 Iter 11848, loss: 0.5505 Iter 11849, loss: 0.4980

Iter 11850, loss: 0.4996 Iter 11851, loss: 0.4604 Iter 11852, loss: 0.5190 Iter 11853, loss: 0.5469 Iter 11854, loss: 0.5133 Iter 11855, loss: 0.5302 Iter 11856, loss: 0.5096 Iter 11857, loss: 0.5408 Iter 11858, loss: 0.5116 Iter 11859, loss: 0.5250 Iter 11860, loss: 0.5081 Iter 11861, loss: 0.5256 Iter 11862, loss: 0.5578 Iter 11863, loss: 0.5087 Iter 11864, loss: 0.4893 Iter 11865, loss: 0.5038 Iter 11866, loss: 0.4941 Iter 11867, loss: 0.4995 Iter 11868, loss: 0.5180 Iter 11869, loss: 0.5139 Iter 11870, loss: 0.4739 Iter 11871, loss: 0.5211 Iter 11872, loss: 0.4628 Iter 11873, loss: 0.4564 Iter 11874, loss: 0.5132 Iter 11875, loss: 0.4882 Iter 11876, loss: 0.5291 Iter 11877, loss: 0.4994 Iter 11878, loss: 0.5231 Iter 11879, loss: 0.5049 Iter 11880, loss: 0.4885 Iter 11881, loss: 0.5020 Iter 11882, loss: 0.5150 Iter 11883, loss: 0.5379 Iter 11884, loss: 0.4585 Iter 11885, loss: 0.4898 Iter 11886, loss: 0.5207 Iter 11887, loss: 0.5136 Iter 11888, loss: 0.4612 Iter 11889, loss: 0.5455 Iter 11890, loss: 0.5112 Iter 11891, loss: 0.5279

Iter 11892, loss: 0.4935 Iter 11893, loss: 0.5338 Iter 11894, loss: 0.4896 Iter 11895, loss: 0.4779 Iter 11896, loss: 0.5183 Iter 11897, loss: 0.5138 Iter 11898, loss: 0.4659 Iter 11899, loss: 0.4604 Iter 11900, loss: 0.5126 Iter 11901, loss: 0.5475 Iter 11902, loss: 0.4709 Iter 11903, loss: 0.5453 Iter 11904, loss: 0.4889 Iter 11905, loss: 0.4665 Iter 11906, loss: 0.4506 Iter 11907, loss: 0.4791 Iter 11908, loss: 0.4971 Iter 11909, loss: 0.5045 Iter 11910, loss: 0.5073 Iter 11911, loss: 0.5222 Iter 11912, loss: 0.4875 Iter 11913, loss: 0.4659 Iter 11914, loss: 0.4898 Iter 11915, loss: 0.4608 Iter 11916, loss: 0.5045 Iter 11917, loss: 0.4779 Iter 11918, loss: 0.4957 Iter 11919, loss: 0.4636 Iter 11920, loss: 0.4963 Iter 11921, loss: 0.5515 Iter 11922, loss: 0.5197 Iter 11923, loss: 0.4963 Iter 11924, loss: 0.5067 Iter 11925, loss: 0.5332 Iter 11926, loss: 0.4908 Iter 11927, loss: 0.4687 Iter 11928, loss: 0.4946 Iter 11929, loss: 0.5120 Iter 11930, loss: 0.5054 Iter 11931, loss: 0.4591 Iter 11932, loss: 0.4842 Iter 11933, loss: 0.5319

Iter 11934, loss: 0.5058 Iter 11935, loss: 0.4761 Iter 11936, loss: 0.4754 Iter 11937, loss: 0.4680 Iter 11938, loss: 0.4114 Iter 11939, loss: 0.4987 Iter 11940, loss: 0.5158 Iter 11941, loss: 0.4756 Iter 11942, loss: 0.5332 Iter 11943, loss: 0.5248 Iter 11944, loss: 0.4977 Iter 11945, loss: 0.4935 Iter 11946, loss: 0.4691 Iter 11947, loss: 0.4555 Iter 11948, loss: 0.4774 Iter 11949, loss: 0.5112 Iter 11950, loss: 0.4866 Iter 11951, loss: 0.4930 Iter 11952, loss: 0.4959 Iter 11953, loss: 0.5499 Iter 11954, loss: 0.5291 Iter 11955, loss: 0.5552 Iter 11956, loss: 0.5292 Iter 11957, loss: 0.4701 Iter 11958, loss: 0.5201 Iter 11959, loss: 0.4818 Iter 11960, loss: 0.5159 Iter 11961, loss: 0.5151 Iter 11962, loss: 0.5016 Iter 11963, loss: 0.5259 Iter 11964, loss: 0.4993 Iter 11965, loss: 0.4700 Iter 11966, loss: 0.4833 Iter 11967, loss: 0.4806 Iter 11968, loss: 0.5098 Iter 11969, loss: 0.4983 Iter 11970, loss: 0.4981 Iter 11971, loss: 0.5171 Iter 11972, loss: 0.4859 Iter 11973, loss: 0.4968 Iter 11974, loss: 0.5131 Iter 11975, loss: 0.5053

Iter 11976, loss: 0.4891 Iter 11977, loss: 0.4798 Iter 11978, loss: 0.5055 Iter 11979, loss: 0.4988 Iter 11980, loss: 0.4752 Iter 11981, loss: 0.5185 Iter 11982, loss: 0.5349 Iter 11983, loss: 0.4944 Iter 11984, loss: 0.5202 Iter 11985, loss: 0.4891 Iter 11986, loss: 0.5379 Iter 11987, loss: 0.5295 Iter 11988, loss: 0.5100 Iter 11989, loss: 0.4920 Iter 11990, loss: 0.5211 Iter 11991, loss: 0.4940 Iter 11992, loss: 0.5060 Iter 11993, loss: 0.5021 Iter 11994, loss: 0.4944 Iter 11995, loss: 0.5046 Iter 11996, loss: 0.5326 Iter 11997, loss: 0.5080 Iter 11998, loss: 0.5140 Iter 11999, loss: 0.5294 Iter 12000, loss: 0.4857 Iter 12001, loss: 0.5021 Iter 12002, loss: 0.4432 Iter 12003, loss: 0.5158 Iter 12004, loss: 0.4840 Iter 12005, loss: 0.5119 Iter 12006, loss: 0.4835 Iter 12007, loss: 0.4491 Iter 12008, loss: 0.5020 Iter 12009, loss: 0.4812 Iter 12010, loss: 0.4941 Iter 12011, loss: 0.5175 Iter 12012, loss: 0.5082 Iter 12013, loss: 0.5004 Iter 12014, loss: 0.5169 Iter 12015, loss: 0.4739 Iter 12016, loss: 0.4721 Iter 12017, loss: 0.5023

Iter 12018, loss: 0.4931 Iter 12019, loss: 0.5567 Iter 12020, loss: 0.5182 Iter 12021, loss: 0.4906 Iter 12022, loss: 0.5027 Iter 12023, loss: 0.4731 Iter 12024, loss: 0.5165 Iter 12025, loss: 0.5379 Iter 12026, loss: 0.4873 Iter 12027, loss: 0.4802 Iter 12028, loss: 0.4975 Iter 12029, loss: 0.4996 Iter 12030, loss: 0.5106 Iter 12031, loss: 0.5274 Iter 12032, loss: 0.4950 Iter 12033, loss: 0.5079 Iter 12034, loss: 0.5055 Iter 12035, loss: 0.5131 Iter 12036, loss: 0.4872 Iter 12037, loss: 0.5248 Iter 12038, loss: 0.5125 Iter 12039, loss: 0.5112 Iter 12040, loss: 0.4916 Iter 12041, loss: 0.4945 Iter 12042, loss: 0.5184 Iter 12043, loss: 0.5284 Iter 12044, loss: 0.5062 Iter 12045, loss: 0.4676 Iter 12046, loss: 0.5648 Iter 12047, loss: 0.4943 Iter 12048, loss: 0.4821 Iter 12049, loss: 0.5202 Iter 12050, loss: 0.4795 Iter 12051, loss: 0.4897 Iter 12052, loss: 0.4801 Iter 12053, loss: 0.4939 Iter 12054, loss: 0.4959 Iter 12055, loss: 0.5330 Iter 12056, loss: 0.5016 Iter 12057, loss: 0.5078 Iter 12058, loss: 0.4926 Iter 12059, loss: 0.4743

Iter 12060, loss: 0.4781 Iter 12061, loss: 0.5253 Iter 12062, loss: 0.4975 Iter 12063, loss: 0.5085 Iter 12064, loss: 0.4940 Iter 12065, loss: 0.5015 Iter 12066, loss: 0.5255 Iter 12067, loss: 0.4803 Iter 12068, loss: 0.5071 Iter 12069, loss: 0.5217 Iter 12070, loss: 0.5151 Iter 12071, loss: 0.5082 Iter 12072, loss: 0.4906 Iter 12073, loss: 0.5077 Iter 12074, loss: 0.5110 Iter 12075, loss: 0.4510 Iter 12076, loss: 0.4776 Iter 12077, loss: 0.4773 Iter 12078, loss: 0.5078 Iter 12079, loss: 0.5053 Iter 12080, loss: 0.5063 Iter 12081, loss: 0.4588 Iter 12082, loss: 0.4651 Iter 12083, loss: 0.5182 Iter 12084, loss: 0.5088 Iter 12085, loss: 0.4750 Iter 12086, loss: 0.4821 Iter 12087, loss: 0.4736 Iter 12088, loss: 0.5337 Iter 12089, loss: 0.4508 Iter 12090, loss: 0.5089 Iter 12091, loss: 0.4650 Iter 12092, loss: 0.5139 Iter 12093, loss: 0.5032 Iter 12094, loss: 0.4649 Iter 12095, loss: 0.5194 Iter 12096, loss: 0.4932 Iter 12097, loss: 0.5353 Iter 12098, loss: 0.5271 Iter 12099, loss: 0.4681 Iter 12100, loss: 0.4986 Iter 12101, loss: 0.4900

Iter 12102, loss: 0.4579 Iter 12103, loss: 0.5146 Iter 12104, loss: 0.4726 Iter 12105, loss: 0.4642 Iter 12106, loss: 0.5089 Iter 12107, loss: 0.5057 Iter 12108, loss: 0.4911 Iter 12109, loss: 0.5182 Iter 12110, loss: 0.5314 Iter 12111, loss: 0.5018 Iter 12112, loss: 0.4726 Iter 12113, loss: 0.5310 Iter 12114, loss: 0.5185 Iter 12115, loss: 0.4795 Iter 12116, loss: 0.5063 Iter 12117, loss: 0.5490 Iter 12118, loss: 0.5092 Iter 12119, loss: 0.4869 Iter 12120, loss: 0.5032 Iter 12121, loss: 0.5173 Iter 12122, loss: 0.5223 Iter 12123, loss: 0.5324 Iter 12124, loss: 0.4746 Iter 12125, loss: 0.4707 Iter 12126, loss: 0.5368 Iter 12127, loss: 0.5136 Iter 12128, loss: 0.5124 Iter 12129, loss: 0.4723 Iter 12130, loss: 0.4930 Iter 12131, loss: 0.4959 Iter 12132, loss: 0.4641 Iter 12133, loss: 0.5082 Iter 12134, loss: 0.4863 Iter 12135, loss: 0.4860 Iter 12136, loss: 0.5084 Iter 12137, loss: 0.4995 Iter 12138, loss: 0.4902 Iter 12139, loss: 0.5299 Iter 12140, loss: 0.4916 Iter 12141, loss: 0.4832 Iter 12142, loss: 0.5306 Iter 12143, loss: 0.4779

Iter 12144, loss: 0.4835 Iter 12145, loss: 0.5193 Iter 12146, loss: 0.5117 Iter 12147, loss: 0.4607 Iter 12148, loss: 0.5158 Iter 12149, loss: 0.5252 Iter 12150, loss: 0.4512 Iter 12151, loss: 0.4536 Iter 12152, loss: 0.4955 Iter 12153, loss: 0.4999 Iter 12154, loss: 0.5035 Iter 12155, loss: 0.4796 Iter 12156, loss: 0.4780 Iter 12157, loss: 0.5526 Iter 12158, loss: 0.4955 Iter 12159, loss: 0.4564 Iter 12160, loss: 0.4655 Iter 12161, loss: 0.5193 Iter 12162, loss: 0.5195 Iter 12163, loss: 0.4872 Iter 12164, loss: 0.5094 Iter 12165, loss: 0.4879 Iter 12166, loss: 0.5415 Iter 12167, loss: 0.5077 Iter 12168, loss: 0.5206 Iter 12169, loss: 0.5152 Iter 12170, loss: 0.4601 Iter 12171, loss: 0.4856 Iter 12172, loss: 0.5354 Iter 12173, loss: 0.4787 Iter 12174, loss: 0.5217 Iter 12175, loss: 0.4721 Iter 12176, loss: 0.5199 Iter 12177, loss: 0.4882 Iter 12178, loss: 0.5084 Iter 12179, loss: 0.5108 Iter 12180, loss: 0.4663 Iter 12181, loss: 0.4772 Iter 12182, loss: 0.5056 Iter 12183, loss: 0.4663 Iter 12184, loss: 0.4819 Iter 12185, loss: 0.4772

Iter 12186, loss: 0.5118 Iter 12187, loss: 0.4908 Iter 12188, loss: 0.5058 Iter 12189, loss: 0.4989 Iter 12190, loss: 0.5179 Iter 12191, loss: 0.4866 Iter 12192, loss: 0.5035 Iter 12193, loss: 0.5020 Iter 12194, loss: 0.5234 Iter 12195, loss: 0.5209 Iter 12196, loss: 0.4967 Iter 12197, loss: 0.5001 Iter 12198, loss: 0.5098 Iter 12199, loss: 0.4890 Iter 12200, loss: 0.4519 Iter 12201, loss: 0.5147 Iter 12202, loss: 0.5467 Iter 12203, loss: 0.4704 Iter 12204, loss: 0.4848 Iter 12205, loss: 0.4737 Iter 12206, loss: 0.5011 Iter 12207, loss: 0.4896 Iter 12208, loss: 0.4843 Iter 12209, loss: 0.5060 Iter 12210, loss: 0.5402 Iter 12211, loss: 0.5241 Iter 12212, loss: 0.4943 Iter 12213, loss: 0.5163 Iter 12214, loss: 0.4682 Iter 12215, loss: 0.4784 Iter 12216, loss: 0.4706 Iter 12217, loss: 0.4886 Iter 12218, loss: 0.4788 Iter 12219, loss: 0.4700 Iter 12220, loss: 0.4915 Iter 12221, loss: 0.4768 Iter 12222, loss: 0.4899 Iter 12223, loss: 0.5027 Iter 12224, loss: 0.4806 Iter 12225, loss: 0.4870 Iter 12226, loss: 0.4824 Iter 12227, loss: 0.4634

Iter 12228, loss: 0.5048 Iter 12229, loss: 0.4861 Iter 12230, loss: 0.4586 Iter 12231, loss: 0.4922 Iter 12232, loss: 0.5192 Iter 12233, loss: 0.4941 Iter 12234, loss: 0.5162 Iter 12235, loss: 0.4566 Iter 12236, loss: 0.4949 Iter 12237, loss: 0.4734 Iter 12238, loss: 0.4981 Iter 12239, loss: 0.5041 Iter 12240, loss: 0.4737 Iter 12241, loss: 0.5251 Iter 12242, loss: 0.5330 Iter 12243, loss: 0.4964 Iter 12244, loss: 0.4762 Iter 12245, loss: 0.4700 Iter 12246, loss: 0.5264 Iter 12247, loss: 0.4847 Iter 12248, loss: 0.4766 Iter 12249, loss: 0.4909 Iter 12250, loss: 0.5152 Iter 12251, loss: 0.5419 Iter 12252, loss: 0.4946 Iter 12253, loss: 0.4634 Iter 12254, loss: 0.5041 Iter 12255, loss: 0.5044 Iter 12256, loss: 0.4915 Iter 12257, loss: 0.5262 Iter 12258, loss: 0.5310 Iter 12259, loss: 0.4720 Iter 12260, loss: 0.4851 Iter 12261, loss: 0.5255 Iter 12262, loss: 0.4962 Iter 12263, loss: 0.4892 Iter 12264, loss: 0.5094 Iter 12265, loss: 0.5292 Iter 12266, loss: 0.5367 Iter 12267, loss: 0.5077 Iter 12268, loss: 0.5084 Iter 12269, loss: 0.5465

Iter 12270, loss: 0.5751 Iter 12271, loss: 0.5136 Iter 12272, loss: 0.4671 Iter 12273, loss: 0.5019 Iter 12274, loss: 0.4888 Iter 12275, loss: 0.5531 Iter 12276, loss: 0.5115 Iter 12277, loss: 0.4933 Iter 12278, loss: 0.4755 Iter 12279, loss: 0.4944 Iter 12280, loss: 0.4734 Iter 12281, loss: 0.5082 Iter 12282, loss: 0.4677 Iter 12283, loss: 0.5710 Iter 12284, loss: 0.5277 Iter 12285, loss: 0.4863 Iter 12286, loss: 0.5358 Iter 12287, loss: 0.5002 Iter 12288, loss: 0.5329 Iter 12289, loss: 0.5344 Iter 12290, loss: 0.5014 Iter 12291, loss: 0.4820 Iter 12292, loss: 0.4981 Iter 12293, loss: 0.4752 Iter 12294, loss: 0.5151 Iter 12295, loss: 0.4927 Iter 12296, loss: 0.5166 Iter 12297, loss: 0.5232 Iter 12298, loss: 0.4831 Iter 12299, loss: 0.5157 Iter 12300, loss: 0.4996 Iter 12301, loss: 0.4962 Iter 12302, loss: 0.5294 Iter 12303, loss: 0.5145 Iter 12304, loss: 0.4551 Iter 12305, loss: 0.5208 Iter 12306, loss: 0.4973 Iter 12307, loss: 0.5075 Iter 12308, loss: 0.4775 Iter 12309, loss: 0.4780 Iter 12310, loss: 0.5029 Iter 12311, loss: 0.5239

Iter 12312, loss: 0.4635 Iter 12313, loss: 0.4876 Iter 12314, loss: 0.5315 Iter 12315, loss: 0.4754 Iter 12316, loss: 0.4798 Iter 12317, loss: 0.5073 Iter 12318, loss: 0.5046 Iter 12319, loss: 0.5350 Iter 12320, loss: 0.4779 Iter 12321, loss: 0.5101 Iter 12322, loss: 0.4750 Iter 12323, loss: 0.4803 Iter 12324, loss: 0.4809 Iter 12325, loss: 0.5277 Iter 12326, loss: 0.5266 Iter 12327, loss: 0.4937 Iter 12328, loss: 0.4688 Iter 12329, loss: 0.4432 Iter 12330, loss: 0.4548 Iter 12331, loss: 0.4692 Iter 12332, loss: 0.4672 Iter 12333, loss: 0.4821 Iter 12334, loss: 0.5067 Iter 12335, loss: 0.4815 Iter 12336, loss: 0.5368 Iter 12337, loss: 0.4660 Iter 12338, loss: 0.4932 Iter 12339, loss: 0.5040 Iter 12340, loss: 0.5297 Iter 12341, loss: 0.4843 Iter 12342, loss: 0.4877 Iter 12343, loss: 0.4773 Iter 12344, loss: 0.4651 Iter 12345, loss: 0.5407 Iter 12346, loss: 0.5144 Iter 12347, loss: 0.5244 Iter 12348, loss: 0.5111 Iter 12349, loss: 0.4794 Iter 12350, loss: 0.5136 Iter 12351, loss: 0.5007 Iter 12352, loss: 0.5084 Iter 12353, loss: 0.4639

Iter 12354, loss: 0.5266 Iter 12355, loss: 0.4882 Iter 12356, loss: 0.4867 Iter 12357, loss: 0.5229 Iter 12358, loss: 0.5027 Iter 12359, loss: 0.4627 Iter 12360, loss: 0.5454 Iter 12361, loss: 0.4844 Iter 12362, loss: 0.4685 Iter 12363, loss: 0.4797 Iter 12364, loss: 0.4776 Iter 12365, loss: 0.5218 Iter 12366, loss: 0.5035 Iter 12367, loss: 0.4455 Iter 12368, loss: 0.4613 Iter 12369, loss: 0.5240 Iter 12370, loss: 0.4771 Iter 12371, loss: 0.4698 Iter 12372, loss: 0.4738 Iter 12373, loss: 0.4922 Iter 12374, loss: 0.4705 Iter 12375, loss: 0.4773 Iter 12376, loss: 0.5111 Iter 12377, loss: 0.5081 Iter 12378, loss: 0.4984 Iter 12379, loss: 0.5189 Iter 12380, loss: 0.5338 Iter 12381, loss: 0.5048 Iter 12382, loss: 0.4819 Iter 12383, loss: 0.4755 Iter 12384, loss: 0.5031 Iter 12385, loss: 0.5030 Iter 12386, loss: 0.5077 Iter 12387, loss: 0.4761 Iter 12388, loss: 0.5026 Iter 12389, loss: 0.4965 Iter 12390, loss: 0.5105 Iter 12391, loss: 0.5209 Iter 12392, loss: 0.5194 Iter 12393, loss: 0.4829 Iter 12394, loss: 0.5073 Iter 12395, loss: 0.5358

Iter 12396, loss: 0.4679 Iter 12397, loss: 0.5308 Iter 12398, loss: 0.5203 Iter 12399, loss: 0.4927 Iter 12400, loss: 0.5048 Iter 12401, loss: 0.4598 Iter 12402, loss: 0.4786 Iter 12403, loss: 0.5008 Iter 12404, loss: 0.5325 Iter 12405, loss: 0.5243 Iter 12406, loss: 0.5207 Iter 12407, loss: 0.4749 Iter 12408, loss: 0.5588 Iter 12409, loss: 0.4641 Iter 12410, loss: 0.4754 Iter 12411, loss: 0.4702 Iter 12412, loss: 0.4844 Iter 12413, loss: 0.5502 Iter 12414, loss: 0.4895 Iter 12415, loss: 0.4710 Iter 12416, loss: 0.4920 Iter 12417, loss: 0.5157 Iter 12418, loss: 0.5236 Iter 12419, loss: 0.5281 Iter 12420, loss: 0.5043 Iter 12421, loss: 0.4921 Iter 12422, loss: 0.5043 Iter 12423, loss: 0.4788 Iter 12424, loss: 0.5615 Iter 12425, loss: 0.5103 Iter 12426, loss: 0.4903 Iter 12427, loss: 0.4814 Iter 12428, loss: 0.4951 Iter 12429, loss: 0.4985 Iter 12430, loss: 0.5467 Iter 12431, loss: 0.4926 Iter 12432, loss: 0.4680 Iter 12433, loss: 0.4949 Iter 12434, loss: 0.5007 Iter 12435, loss: 0.5173 Iter 12436, loss: 0.4995 Iter 12437, loss: 0.4775

Iter 12438, loss: 0.5302 Iter 12439, loss: 0.4761 Iter 12440, loss: 0.5074 Iter 12441, loss: 0.4690 Iter 12442, loss: 0.5423 Iter 12443, loss: 0.4628 Iter 12444, loss: 0.5200 Iter 12445, loss: 0.4934 Iter 12446, loss: 0.4619 Iter 12447, loss: 0.4638 Iter 12448, loss: 0.4882 Iter 12449, loss: 0.4832 Iter 12450, loss: 0.5405 Iter 12451, loss: 0.5086 Iter 12452, loss: 0.5325 Iter 12453, loss: 0.4931 Iter 12454, loss: 0.5353 Iter 12455, loss: 0.4890 Iter 12456, loss: 0.4898 Iter 12457, loss: 0.4861 Iter 12458, loss: 0.4768 Iter 12459, loss: 0.4960 Iter 12460, loss: 0.4995 Iter 12461, loss: 0.5348 Iter 12462, loss: 0.5103 Iter 12463, loss: 0.5509 Iter 12464, loss: 0.4967 Iter 12465, loss: 0.5279 Iter 12466, loss: 0.5322 Iter 12467, loss: 0.4926 Iter 12468, loss: 0.4991 Iter 12469, loss: 0.5268 Iter 12470, loss: 0.4610 Iter 12471, loss: 0.5100 Iter 12472, loss: 0.5160 Iter 12473, loss: 0.4667 Iter 12474, loss: 0.4948 Iter 12475, loss: 0.4900 Iter 12476, loss: 0.4922 Iter 12477, loss: 0.4869 Iter 12478, loss: 0.4957 Iter 12479, loss: 0.4913

Iter 12480, loss: 0.4795 Iter 12481, loss: 0.5382 Iter 12482, loss: 0.5138 Iter 12483, loss: 0.5196 Iter 12484, loss: 0.5093 Iter 12485, loss: 0.5036 Iter 12486, loss: 0.4801 Iter 12487, loss: 0.5298 Iter 12488, loss: 0.5313 Iter 12489, loss: 0.4885 Iter 12490, loss: 0.5288 Iter 12491, loss: 0.4862 Iter 12492, loss: 0.4959 Iter 12493, loss: 0.4490 Iter 12494, loss: 0.4790 Iter 12495, loss: 0.5355 Iter 12496, loss: 0.5002 Iter 12497, loss: 0.4719 Iter 12498, loss: 0.4657 Iter 12499, loss: 0.4978 Iter 12500, loss: 0.5056 Iter 12501, loss: 0.4974 Iter 12502, loss: 0.5222 Iter 12503, loss: 0.4896 Iter 12504, loss: 0.5158 Iter 12505, loss: 0.4870 Iter 12506, loss: 0.5108 Iter 12507, loss: 0.5326 Iter 12508, loss: 0.5607 Iter 12509, loss: 0.5006 Iter 12510, loss: 0.4877 Iter 12511, loss: 0.5067 Iter 12512, loss: 0.4562 Iter 12513, loss: 0.5020 Iter 12514, loss: 0.5084 Iter 12515, loss: 0.4786 Iter 12516, loss: 0.5267 Iter 12517, loss: 0.5120 Iter 12518, loss: 0.5073 Iter 12519, loss: 0.4684 Iter 12520, loss: 0.5286 Iter 12521, loss: 0.4986

Iter 12522, loss: 0.4951 Iter 12523, loss: 0.4810 Iter 12524, loss: 0.4807 Iter 12525, loss: 0.4762 Iter 12526, loss: 0.4810 Iter 12527, loss: 0.4791 Iter 12528, loss: 0.5350 Iter 12529, loss: 0.4993 Iter 12530, loss: 0.4574 Iter 12531, loss: 0.5004 Iter 12532, loss: 0.5055 Iter 12533, loss: 0.5124 Iter 12534, loss: 0.5006 Iter 12535, loss: 0.4673 Iter 12536, loss: 0.4785 Iter 12537, loss: 0.4996 Iter 12538, loss: 0.5196 Iter 12539, loss: 0.5346 Iter 12540, loss: 0.5139 Iter 12541, loss: 0.5004 Iter 12542, loss: 0.4661 Iter 12543, loss: 0.4932 Iter 12544, loss: 0.5114 Iter 12545, loss: 0.4619 Iter 12546, loss: 0.5020 Iter 12547, loss: 0.4792 Iter 12548, loss: 0.4753 Iter 12549, loss: 0.4687 Iter 12550, loss: 0.4864 Iter 12551, loss: 0.5008 Iter 12552, loss: 0.5128 Iter 12553, loss: 0.4940 Iter 12554, loss: 0.4667 Iter 12555, loss: 0.4853 Iter 12556, loss: 0.5290 Iter 12557, loss: 0.4906 Iter 12558, loss: 0.5412 Iter 12559, loss: 0.4884 Iter 12560, loss: 0.5252 Iter 12561, loss: 0.4897 Iter 12562, loss: 0.4981 Iter 12563, loss: 0.4891

Iter 12564, loss: 0.4873 Iter 12565, loss: 0.5038 Iter 12566, loss: 0.5074 Iter 12567, loss: 0.4922 Iter 12568, loss: 0.4905 Iter 12569, loss: 0.5157 Iter 12570, loss: 0.5262 Iter 12571, loss: 0.5219 Iter 12572, loss: 0.4933 Iter 12573, loss: 0.5309 Iter 12574, loss: 0.5226 Iter 12575, loss: 0.4937 Iter 12576, loss: 0.4938 Iter 12577, loss: 0.4519 Iter 12578, loss: 0.4945 Iter 12579, loss: 0.5194 Iter 12580, loss: 0.4796 Iter 12581, loss: 0.5285 Iter 12582, loss: 0.5131 Iter 12583, loss: 0.4794 Iter 12584, loss: 0.5152 Iter 12585, loss: 0.5080 Iter 12586, loss: 0.4538 Iter 12587, loss: 0.5240 Iter 12588, loss: 0.4926 Iter 12589, loss: 0.5214 Iter 12590, loss: 0.5306 Iter 12591, loss: 0.4684 Iter 12592, loss: 0.4781 Iter 12593, loss: 0.5334 Iter 12594, loss: 0.5165 Iter 12595, loss: 0.5181 Iter 12596, loss: 0.4714 Iter 12597, loss: 0.5256 Iter 12598, loss: 0.4942 Iter 12599, loss: 0.5093 Iter 12600, loss: 0.4933 Iter 12601, loss: 0.4643 Iter 12602, loss: 0.4855 Iter 12603, loss: 0.4692 Iter 12604, loss: 0.4863 Iter 12605, loss: 0.5053

Iter 12606, loss: 0.5261 Iter 12607, loss: 0.5146 Iter 12608, loss: 0.5059 Iter 12609, loss: 0.5072 Iter 12610, loss: 0.4668 Iter 12611, loss: 0.5112 Iter 12612, loss: 0.4698 Iter 12613, loss: 0.4864 Iter 12614, loss: 0.4812 Iter 12615, loss: 0.5213 Iter 12616, loss: 0.4479 Iter 12617, loss: 0.4842 Iter 12618, loss: 0.4919 Iter 12619, loss: 0.4878 Iter 12620, loss: 0.4499 Iter 12621, loss: 0.4983 Iter 12622, loss: 0.4793 Iter 12623, loss: 0.4782 Iter 12624, loss: 0.5391 Iter 12625, loss: 0.4953 Iter 12626, loss: 0.5357 Iter 12627, loss: 0.4760 Iter 12628, loss: 0.4886 Iter 12629, loss: 0.5109 Iter 12630, loss: 0.4809 Iter 12631, loss: 0.4963 Iter 12632, loss: 0.4778 Iter 12633, loss: 0.5099 Iter 12634, loss: 0.5431 Iter 12635, loss: 0.4698 Iter 12636, loss: 0.4926 Iter 12637, loss: 0.5038 Iter 12638, loss: 0.5219 Iter 12639, loss: 0.5100 Iter 12640, loss: 0.4895 Iter 12641, loss: 0.4807 Iter 12642, loss: 0.5048 Iter 12643, loss: 0.5026 Iter 12644, loss: 0.4972 Iter 12645, loss: 0.4901 Iter 12646, loss: 0.4937 Iter 12647, loss: 0.5080

Iter 12648, loss: 0.4883 Iter 12649, loss: 0.4780 Iter 12650, loss: 0.4811 Iter 12651, loss: 0.4816 Iter 12652, loss: 0.4676 Iter 12653, loss: 0.5232 Iter 12654, loss: 0.5329 Iter 12655, loss: 0.5228 Iter 12656, loss: 0.5097 Iter 12657, loss: 0.4522 Iter 12658, loss: 0.5228 Iter 12659, loss: 0.4668 Iter 12660, loss: 0.5209 Iter 12661, loss: 0.4912 Iter 12662, loss: 0.4799 Iter 12663, loss: 0.4814 Iter 12664, loss: 0.5116 Iter 12665, loss: 0.5176 Iter 12666, loss: 0.5126 Iter 12667, loss: 0.5061 Iter 12668, loss: 0.5359 Iter 12669, loss: 0.4998 Iter 12670, loss: 0.4795 Iter 12671, loss: 0.4870 Iter 12672, loss: 0.5303 Iter 12673, loss: 0.4442 Iter 12674, loss: 0.4964 Iter 12675, loss: 0.4826 Iter 12676, loss: 0.4716 Iter 12677, loss: 0.4880 Iter 12678, loss: 0.5218 Iter 12679, loss: 0.5324 Iter 12680, loss: 0.4890 Iter 12681, loss: 0.5164 Iter 12682, loss: 0.4898 Iter 12683, loss: 0.5106 Iter 12684, loss: 0.5081 Iter 12685, loss: 0.5101 Iter 12686, loss: 0.5429 Iter 12687, loss: 0.4876 Iter 12688, loss: 0.4960 Iter 12689, loss: 0.5051

Iter 12690, loss: 0.5004 Iter 12691, loss: 0.4905 Iter 12692, loss: 0.4691 Iter 12693, loss: 0.4796 Iter 12694, loss: 0.4781 Iter 12695, loss: 0.4520 Iter 12696, loss: 0.4966 Iter 12697, loss: 0.4919 Iter 12698, loss: 0.4973 Iter 12699, loss: 0.4928 Iter 12700, loss: 0.5078 Iter 12701, loss: 0.5260 Iter 12702, loss: 0.5440 Iter 12703, loss: 0.5368 Iter 12704, loss: 0.4969 Iter 12705, loss: 0.4894 Iter 12706, loss: 0.5131 Iter 12707, loss: 0.4819 Iter 12708, loss: 0.4932 Iter 12709, loss: 0.5131 Iter 12710, loss: 0.4978 Iter 12711, loss: 0.4790 Iter 12712, loss: 0.4885 Iter 12713, loss: 0.5033 Iter 12714, loss: 0.4850 Iter 12715, loss: 0.5091 Iter 12716, loss: 0.5079 Iter 12717, loss: 0.5021 Iter 12718, loss: 0.5197 Iter 12719, loss: 0.5008 Iter 12720, loss: 0.4701 Iter 12721, loss: 0.5080 Iter 12722, loss: 0.4721 Iter 12723, loss: 0.4969 Iter 12724, loss: 0.4977 Iter 12725, loss: 0.4613 Iter 12726, loss: 0.4955 Iter 12727, loss: 0.4943 Iter 12728, loss: 0.4579 Iter 12729, loss: 0.5407 Iter 12730, loss: 0.4881 Iter 12731, loss: 0.4933

Iter 12732, loss: 0.4892 Iter 12733, loss: 0.4715 Iter 12734, loss: 0.5035 Iter 12735, loss: 0.5074 Iter 12736, loss: 0.4903 Iter 12737, loss: 0.4952 Iter 12738, loss: 0.5421 Iter 12739, loss: 0.5144 Iter 12740, loss: 0.5137 Iter 12741, loss: 0.5184 Iter 12742, loss: 0.4924 Iter 12743, loss: 0.4497 Iter 12744, loss: 0.5338 Iter 12745, loss: 0.5384 Iter 12746, loss: 0.4882 Iter 12747, loss: 0.4748 Iter 12748, loss: 0.4864 Iter 12749, loss: 0.4968 Iter 12750, loss: 0.4872 Iter 12751, loss: 0.5120 Iter 12752, loss: 0.5394 Iter 12753, loss: 0.4840 Iter 12754, loss: 0.5048 Iter 12755, loss: 0.5324 Iter 12756, loss: 0.5315 Iter 12757, loss: 0.4590 Iter 12758, loss: 0.4675 Iter 12759, loss: 0.4808 Iter 12760, loss: 0.4890 Iter 12761, loss: 0.4851 Iter 12762, loss: 0.4870 Iter 12763, loss: 0.5059 Iter 12764, loss: 0.4715 Iter 12765, loss: 0.4830 Iter 12766, loss: 0.4651 Iter 12767, loss: 0.5167 Iter 12768, loss: 0.4821 Iter 12769, loss: 0.4933 Iter 12770, loss: 0.4834 Iter 12771, loss: 0.4839 Iter 12772, loss: 0.5014 Iter 12773, loss: 0.5128

Iter 12774, loss: 0.5272 Iter 12775, loss: 0.4856 Iter 12776, loss: 0.4818 Iter 12777, loss: 0.4953 Iter 12778, loss: 0.4770 Iter 12779, loss: 0.4667 Iter 12780, loss: 0.4656 Iter 12781, loss: 0.4832 Iter 12782, loss: 0.4932 Iter 12783, loss: 0.4751 Iter 12784, loss: 0.5151 Iter 12785, loss: 0.4650 Iter 12786, loss: 0.5260 Iter 12787, loss: 0.4827 Iter 12788, loss: 0.5155 Iter 12789, loss: 0.4680 Iter 12790, loss: 0.5059 Iter 12791, loss: 0.5125 Iter 12792, loss: 0.4949 Iter 12793, loss: 0.4759 Iter 12794, loss: 0.5098 Iter 12795, loss: 0.4802 Iter 12796, loss: 0.5137 Iter 12797, loss: 0.5180 Iter 12798, loss: 0.5062 Iter 12799, loss: 0.4656 Iter 12800, loss: 0.5004 Iter 12801, loss: 0.4876 Iter 12802, loss: 0.4975 Iter 12803, loss: 0.4901 Iter 12804, loss: 0.5324 Iter 12805, loss: 0.4843 Iter 12806, loss: 0.5255 Iter 12807, loss: 0.4813 Iter 12808, loss: 0.5135 Iter 12809, loss: 0.5012 Iter 12810, loss: 0.5077 Iter 12811, loss: 0.5049 Iter 12812, loss: 0.5099 Iter 12813, loss: 0.4641 Iter 12814, loss: 0.4843 Iter 12815, loss: 0.4697

Iter 12816, loss: 0.5468 Iter 12817, loss: 0.5216 Iter 12818, loss: 0.5016 Iter 12819, loss: 0.4724 Iter 12820, loss: 0.4890 Iter 12821, loss: 0.4880 Iter 12822, loss: 0.4835 Iter 12823, loss: 0.4836 Iter 12824, loss: 0.5149 Iter 12825, loss: 0.4878 Iter 12826, loss: 0.4988 Iter 12827, loss: 0.4554 Iter 12828, loss: 0.4670 Iter 12829, loss: 0.5601 Iter 12830, loss: 0.4931 Iter 12831, loss: 0.4998 Iter 12832, loss: 0.5261 Iter 12833, loss: 0.4733 Iter 12834, loss: 0.5424 Iter 12835, loss: 0.5145 Iter 12836, loss: 0.5175 Iter 12837, loss: 0.4724 Iter 12838, loss: 0.4975 Iter 12839, loss: 0.4898 Iter 12840, loss: 0.4732 Iter 12841, loss: 0.4813 Iter 12842, loss: 0.4754 Iter 12843, loss: 0.4684 Iter 12844, loss: 0.5038 Iter 12845, loss: 0.4797 Iter 12846, loss: 0.5199 Iter 12847, loss: 0.4681 Iter 12848, loss: 0.4857 Iter 12849, loss: 0.5258 Iter 12850, loss: 0.4753 Iter 12851, loss: 0.4877 Iter 12852, loss: 0.4874 Iter 12853, loss: 0.4951 Iter 12854, loss: 0.5088 Iter 12855, loss: 0.4934 Iter 12856, loss: 0.4764 Iter 12857, loss: 0.4988

Iter 12858, loss: 0.5212 Iter 12859, loss: 0.4902 Iter 12860, loss: 0.4963 Iter 12861, loss: 0.5385 Iter 12862, loss: 0.4967 Iter 12863, loss: 0.5121 Iter 12864, loss: 0.4824 Iter 12865, loss: 0.4810 Iter 12866, loss: 0.5435 Iter 12867, loss: 0.5295 Iter 12868, loss: 0.4759 Iter 12869, loss: 0.4845 Iter 12870, loss: 0.5155 Iter 12871, loss: 0.5266 Iter 12872, loss: 0.5239 Iter 12873, loss: 0.4675 Iter 12874, loss: 0.4796 Iter 12875, loss: 0.4734 Iter 12876, loss: 0.5112 Iter 12877, loss: 0.4811 Iter 12878, loss: 0.4748 Iter 12879, loss: 0.4442 Iter 12880, loss: 0.4797 Iter 12881, loss: 0.5377 Iter 12882, loss: 0.5120 Iter 12883, loss: 0.4640 Iter 12884, loss: 0.4907 Iter 12885, loss: 0.4856 Iter 12886, loss: 0.4883 Iter 12887, loss: 0.4824 Iter 12888, loss: 0.4726 Iter 12889, loss: 0.5606 Iter 12890, loss: 0.5089 Iter 12891, loss: 0.4963 Iter 12892, loss: 0.4828 Iter 12893, loss: 0.5165 Iter 12894, loss: 0.4761 Iter 12895, loss: 0.4897 Iter 12896, loss: 0.4933 Iter 12897, loss: 0.4867 Iter 12898, loss: 0.4934 Iter 12899, loss: 0.4827

Iter 12900, loss: 0.4668 Iter 12901, loss: 0.4943 Iter 12902, loss: 0.5204 Iter 12903, loss: 0.5003 Iter 12904, loss: 0.5081 Iter 12905, loss: 0.4995 Iter 12906, loss: 0.5066 Iter 12907, loss: 0.4970 Iter 12908, loss: 0.4851 Iter 12909, loss: 0.4684 Iter 12910, loss: 0.4928 Iter 12911, loss: 0.4793 Iter 12912, loss: 0.5352 Iter 12913, loss: 0.4934 Iter 12914, loss: 0.5042 Iter 12915, loss: 0.5012 Iter 12916, loss: 0.4807 Iter 12917, loss: 0.5134 Iter 12918, loss: 0.4662 Iter 12919, loss: 0.5385 Iter 12920, loss: 0.4861 Iter 12921, loss: 0.4711 Iter 12922, loss: 0.4884 Iter 12923, loss: 0.5167 Iter 12924, loss: 0.4723 Iter 12925, loss: 0.4650 Iter 12926, loss: 0.5110 Iter 12927, loss: 0.4843 Iter 12928, loss: 0.4785 Iter 12929, loss: 0.5249 Iter 12930, loss: 0.5470 Iter 12931, loss: 0.4758 Iter 12932, loss: 0.4925 Iter 12933, loss: 0.4667 Iter 12934, loss: 0.5095 Iter 12935, loss: 0.5056 Iter 12936, loss: 0.5038 Iter 12937, loss: 0.5153 Iter 12938, loss: 0.5184 Iter 12939, loss: 0.4719 Iter 12940, loss: 0.5150 Iter 12941, loss: 0.4988

Iter 12942, loss: 0.4834 Iter 12943, loss: 0.4962 Iter 12944, loss: 0.4559 Iter 12945, loss: 0.4410 Iter 12946, loss: 0.4968 Iter 12947, loss: 0.4997 Iter 12948, loss: 0.4485 Iter 12949, loss: 0.4889 Iter 12950, loss: 0.5046 Iter 12951, loss: 0.4664 Iter 12952, loss: 0.4685 Iter 12953, loss: 0.4848 Iter 12954, loss: 0.5029 Iter 12955, loss: 0.5189 Iter 12956, loss: 0.4931 Iter 12957, loss: 0.4661 Iter 12958, loss: 0.4951 Iter 12959, loss: 0.4985 Iter 12960, loss: 0.4606 Iter 12961, loss: 0.4918 Iter 12962, loss: 0.5075 Iter 12963, loss: 0.5071 Iter 12964, loss: 0.5191 Iter 12965, loss: 0.4636 Iter 12966, loss: 0.5388 Iter 12967, loss: 0.4615 Iter 12968, loss: 0.4889 Iter 12969, loss: 0.5082 Iter 12970, loss: 0.4964 Iter 12971, loss: 0.4773 Iter 12972, loss: 0.4876 Iter 12973, loss: 0.5036 Iter 12974, loss: 0.4758 Iter 12975, loss: 0.5094 Iter 12976, loss: 0.5028 Iter 12977, loss: 0.4934 Iter 12978, loss: 0.4646 Iter 12979, loss: 0.4737 Iter 12980, loss: 0.5053 Iter 12981, loss: 0.5155 Iter 12982, loss: 0.4848

Iter 12983, loss: 0.4991 Iter 12984, loss: 0.4335 Iter 12985, loss: 0.4691 Iter 12986, loss: 0.4764 Iter 12987, loss: 0.4949 Iter 12988, loss: 0.4843 Iter 12989, loss: 0.4933 Iter 12990, loss: 0.4906 Iter 12991, loss: 0.4939 Iter 12992, loss: 0.4861 Iter 12993, loss: 0.4904 Iter 12994, loss: 0.4871 Iter 12995, loss: 0.4713 Iter 12996, loss: 0.4883 Iter 12997, loss: 0.4987 Iter 12998, loss: 0.4863 Iter 12999, loss: 0.5570 Iter 13000, loss: 0.4526 Iter 13001, loss: 0.5158 Iter 13002, loss: 0.4610 Iter 13003, loss: 0.4885 Iter 13004, loss: 0.4580 Iter 13005, loss: 0.4669 Iter 13006, loss: 0.5185 Iter 13007, loss: 0.5145 Iter 13008, loss: 0.5074 Iter 13009, loss: 0.5053 Iter 13010, loss: 0.4777 Iter 13011, loss: 0.4765 Iter 13012, loss: 0.4741 Iter 13013, loss: 0.4796 Iter 13014, loss: 0.4930 Iter 13015, loss: 0.4735 Iter 13016, loss: 0.5078 Iter 13017, loss: 0.4907 Iter 13018, loss: 0.5199 Iter 13019, loss: 0.5103 Iter 13020, loss: 0.5125 Iter 13021, loss: 0.4456 Iter 13022, loss: 0.5318 Iter 13023, loss: 0.5074 Iter 13024, loss: 0.4738

Iter 13025, loss: 0.5112 Iter 13026, loss: 0.4257 Iter 13027, loss: 0.4990 Iter 13028, loss: 0.4937 Iter 13029, loss: 0.4945 Iter 13030, loss: 0.4708 Iter 13031, loss: 0.4814 Iter 13032, loss: 0.4923 Iter 13033, loss: 0.4739 Iter 13034, loss: 0.5325 Iter 13035, loss: 0.4519 Iter 13036, loss: 0.4856 Iter 13037, loss: 0.5146 Iter 13038, loss: 0.4859 Iter 13039, loss: 0.4836 Iter 13040, loss: 0.5248 Iter 13041, loss: 0.5148 Iter 13042, loss: 0.4727 Iter 13043, loss: 0.5048 Iter 13044, loss: 0.4802 Iter 13045, loss: 0.4933 Iter 13046, loss: 0.5122 Iter 13047, loss: 0.5011 Iter 13048, loss: 0.5137 Iter 13049, loss: 0.5450 Iter 13050, loss: 0.5015 Iter 13051, loss: 0.4885 Iter 13052, loss: 0.4483 Iter 13053, loss: 0.5147 Iter 13054, loss: 0.4958 Iter 13055, loss: 0.4910 Iter 13056, loss: 0.5109 Iter 13057, loss: 0.4462 Iter 13058, loss: 0.4392 Iter 13059, loss: 0.5249 Iter 13060, loss: 0.4853 Iter 13061, loss: 0.5117 Iter 13062, loss: 0.4803 Iter 13063, loss: 0.4795 Iter 13064, loss: 0.5380 Iter 13065, loss: 0.4777 Iter 13066, loss: 0.4986

Iter 13067, loss: 0.5160 Iter 13068, loss: 0.4755 Iter 13069, loss: 0.4713 Iter 13070, loss: 0.5292 Iter 13071, loss: 0.5323 Iter 13072, loss: 0.5029 Iter 13073, loss: 0.5139 Iter 13074, loss: 0.5220 Iter 13075, loss: 0.4958 Iter 13076, loss: 0.5285 Iter 13077, loss: 0.5056 Iter 13078, loss: 0.4560 Iter 13079, loss: 0.5234 Iter 13080, loss: 0.4981 Iter 13081, loss: 0.4949 Iter 13082, loss: 0.4515 Iter 13083, loss: 0.5151 Iter 13084, loss: 0.5201 Iter 13085, loss: 0.4874 Iter 13086, loss: 0.5146 Iter 13087, loss: 0.4898 Iter 13088, loss: 0.4697 Iter 13089, loss: 0.4665 Iter 13090, loss: 0.4792 Iter 13091, loss: 0.5149 Iter 13092, loss: 0.4887 Iter 13093, loss: 0.4591 Iter 13094, loss: 0.4925 Iter 13095, loss: 0.5210 Iter 13096, loss: 0.5118 Iter 13097, loss: 0.5062 Iter 13098, loss: 0.4857 Iter 13099, loss: 0.4766 Iter 13100, loss: 0.4941 Iter 13101, loss: 0.4938 Iter 13102, loss: 0.4630 Iter 13103, loss: 0.5056 Iter 13104, loss: 0.4984 Iter 13105, loss: 0.4969 Iter 13106, loss: 0.4998 Iter 13107, loss: 0.5223 Iter 13108, loss: 0.5225

Iter 13109, loss: 0.4794 Iter 13110, loss: 0.4692 Iter 13111, loss: 0.5046 Iter 13112, loss: 0.4902 Iter 13113, loss: 0.4492 Iter 13114, loss: 0.4535 Iter 13115, loss: 0.5267 Iter 13116, loss: 0.4875 Iter 13117, loss: 0.4693 Iter 13118, loss: 0.5462 Iter 13119, loss: 0.4974 Iter 13120, loss: 0.4479 Iter 13121, loss: 0.5155 Iter 13122, loss: 0.4785 Iter 13123, loss: 0.4606 Iter 13124, loss: 0.5080 Iter 13125, loss: 0.4837 Iter 13126, loss: 0.4713 Iter 13127, loss: 0.4592 Iter 13128, loss: 0.4712 Iter 13129, loss: 0.4619 Iter 13130, loss: 0.4623 Iter 13131, loss: 0.4970 Iter 13132, loss: 0.4936 Iter 13133, loss: 0.4915 Iter 13134, loss: 0.4679 Iter 13135, loss: 0.4651 Iter 13136, loss: 0.5027 Iter 13137, loss: 0.4579 Iter 13138, loss: 0.4735 Iter 13139, loss: 0.5242 Iter 13140, loss: 0.5300 Iter 13141, loss: 0.4782 Iter 13142, loss: 0.4967 Iter 13143, loss: 0.5146 Iter 13144, loss: 0.5094 Iter 13145, loss: 0.4823 Iter 13146, loss: 0.5152 Iter 13147, loss: 0.5192 Iter 13148, loss: 0.5154 Iter 13149, loss: 0.5078 Iter 13150, loss: 0.4620

Iter 13151, loss: 0.4994 Iter 13152, loss: 0.5288 Iter 13153, loss: 0.4918 Iter 13154, loss: 0.5402 Iter 13155, loss: 0.5253 Iter 13156, loss: 0.4777 Iter 13157, loss: 0.4959 Iter 13158, loss: 0.5143 Iter 13159, loss: 0.5048 Iter 13160, loss: 0.4966 Iter 13161, loss: 0.4965 Iter 13162, loss: 0.5225 Iter 13163, loss: 0.4560 Iter 13164, loss: 0.4810 Iter 13165, loss: 0.5012 Iter 13166, loss: 0.4856 Iter 13167, loss: 0.4932 Iter 13168, loss: 0.4808 Iter 13169, loss: 0.5328 Iter 13170, loss: 0.4909 Iter 13171, loss: 0.4858 Iter 13172, loss: 0.5280 Iter 13173, loss: 0.4617 Iter 13174, loss: 0.5035 Iter 13175, loss: 0.5027 Iter 13176, loss: 0.4997 Iter 13177, loss: 0.5140 Iter 13178, loss: 0.4745 Iter 13179, loss: 0.5114 Iter 13180, loss: 0.4705 Iter 13181, loss: 0.4796 Iter 13182, loss: 0.4652 Iter 13183, loss: 0.4788 Iter 13184, loss: 0.4573 Iter 13185, loss: 0.4875 Iter 13186, loss: 0.5205 Iter 13187, loss: 0.4839 Iter 13188, loss: 0.4906 Iter 13189, loss: 0.4573 Iter 13190, loss: 0.4835 Iter 13191, loss: 0.5125 Iter 13192, loss: 0.4780

Iter 13193, loss: 0.5123 Iter 13194, loss: 0.5188 Iter 13195, loss: 0.5138 Iter 13196, loss: 0.4668 Iter 13197, loss: 0.4904 Iter 13198, loss: 0.4941 Iter 13199, loss: 0.4838 Iter 13200, loss: 0.5012 Iter 13201, loss: 0.5164 Iter 13202, loss: 0.4668 Iter 13203, loss: 0.5220 Iter 13204, loss: 0.4639 Iter 13205, loss: 0.4690 Iter 13206, loss: 0.5058 Iter 13207, loss: 0.4224 Iter 13208, loss: 0.4790 Iter 13209, loss: 0.5138 Iter 13210, loss: 0.4960 Iter 13211, loss: 0.4954 Iter 13212, loss: 0.5117 Iter 13213, loss: 0.5319 Iter 13214, loss: 0.4845 Iter 13215, loss: 0.5087 Iter 13216, loss: 0.5217 Iter 13217, loss: 0.4366 Iter 13218, loss: 0.5027 Iter 13219, loss: 0.5184 Iter 13220, loss: 0.5186 Iter 13221, loss: 0.5031 Iter 13222, loss: 0.4993 Iter 13223, loss: 0.5183 Iter 13224, loss: 0.4703 Iter 13225, loss: 0.4938 Iter 13226, loss: 0.4897 Iter 13227, loss: 0.5331 Iter 13228, loss: 0.4953 Iter 13229, loss: 0.4741 Iter 13230, loss: 0.4958 Iter 13231, loss: 0.4920 Iter 13232, loss: 0.4989 Iter 13233, loss: 0.5035 Iter 13234, loss: 0.4721

Iter 13235, loss: 0.4750 Iter 13236, loss: 0.5099 Iter 13237, loss: 0.4544 Iter 13238, loss: 0.5012 Iter 13239, loss: 0.5199 Iter 13240, loss: 0.4540 Iter 13241, loss: 0.4831 Iter 13242, loss: 0.4920 Iter 13243, loss: 0.4715 Iter 13244, loss: 0.4891 Iter 13245, loss: 0.4867 Iter 13246, loss: 0.5204 Iter 13247, loss: 0.4510 Iter 13248, loss: 0.5187 Iter 13249, loss: 0.4746 Iter 13250, loss: 0.5079 Iter 13251, loss: 0.5306 Iter 13252, loss: 0.5055 Iter 13253, loss: 0.5350 Iter 13254, loss: 0.4729 Iter 13255, loss: 0.5010 Iter 13256, loss: 0.4494 Iter 13257, loss: 0.4873 Iter 13258, loss: 0.5201 Iter 13259, loss: 0.5133 Iter 13260, loss: 0.4862 Iter 13261, loss: 0.5191 Iter 13262, loss: 0.4823 Iter 13263, loss: 0.5001 Iter 13264, loss: 0.4725 Iter 13265, loss: 0.4994 Iter 13266, loss: 0.5044 Iter 13267, loss: 0.4834 Iter 13268, loss: 0.5256 Iter 13269, loss: 0.4635 Iter 13270, loss: 0.4635 Iter 13271, loss: 0.5082 Iter 13272, loss: 0.5113 Iter 13273, loss: 0.4844 Iter 13274, loss: 0.5021 Iter 13275, loss: 0.5538 Iter 13276, loss: 0.5009

Iter 13277, loss: 0.5240 Iter 13278, loss: 0.4861 Iter 13279, loss: 0.5231 Iter 13280, loss: 0.4674 Iter 13281, loss: 0.4750 Iter 13282, loss: 0.4704 Iter 13283, loss: 0.4726 Iter 13284, loss: 0.4998 Iter 13285, loss: 0.4509 Iter 13286, loss: 0.4910 Iter 13287, loss: 0.5165 Iter 13288, loss: 0.5134 Iter 13289, loss: 0.5135 Iter 13290, loss: 0.4876 Iter 13291, loss: 0.5090 Iter 13292, loss: 0.4788 Iter 13293, loss: 0.4848 Iter 13294, loss: 0.5099 Iter 13295, loss: 0.4887 Iter 13296, loss: 0.5307 Iter 13297, loss: 0.4531 Iter 13298, loss: 0.5260 Iter 13299, loss: 0.4907 Iter 13300, loss: 0.5122 Iter 13301, loss: 0.4547 Iter 13302, loss: 0.4803 Iter 13303, loss: 0.4837 Iter 13304, loss: 0.4561 Iter 13305, loss: 0.5267 Iter 13306, loss: 0.4947 Iter 13307, loss: 0.4594 Iter 13308, loss: 0.5175 Iter 13309, loss: 0.4617 Iter 13310, loss: 0.5266 Iter 13311, loss: 0.4852 Iter 13312, loss: 0.4525 Iter 13313, loss: 0.4973 Iter 13314, loss: 0.4879 Iter 13315, loss: 0.4705 Iter 13316, loss: 0.5099 Iter 13317, loss: 0.5344 Iter 13318, loss: 0.4689

Iter 13319, loss: 0.4870 Iter 13320, loss: 0.5186 Iter 13321, loss: 0.4613 Iter 13322, loss: 0.5041 Iter 13323, loss: 0.5454 Iter 13324, loss: 0.4804 Iter 13325, loss: 0.4714 Iter 13326, loss: 0.4804 Iter 13327, loss: 0.5069 Iter 13328, loss: 0.5244 Iter 13329, loss: 0.4720 Iter 13330, loss: 0.5164 Iter 13331, loss: 0.5282 Iter 13332, loss: 0.4863 Iter 13333, loss: 0.4794 Iter 13334, loss: 0.4563 Iter 13335, loss: 0.5367 Iter 13336, loss: 0.4982 Iter 13337, loss: 0.4914 Iter 13338, loss: 0.5201 Iter 13339, loss: 0.4797 Iter 13340, loss: 0.4821 Iter 13341, loss: 0.4836 Iter 13342, loss: 0.5172 Iter 13343, loss: 0.4598 Iter 13344, loss: 0.4628 Iter 13345, loss: 0.4666 Iter 13346, loss: 0.4916 Iter 13347, loss: 0.4992 Iter 13348, loss: 0.5154 Iter 13349, loss: 0.4967 Iter 13350, loss: 0.5021 Iter 13351, loss: 0.4820 Iter 13352, loss: 0.4747 Iter 13353, loss: 0.4538 Iter 13354, loss: 0.4926 Iter 13355, loss: 0.4664 Iter 13356, loss: 0.4748 Iter 13357, loss: 0.4934 Iter 13358, loss: 0.4966 Iter 13359, loss: 0.5046 Iter 13360, loss: 0.5119

Iter 13361, loss: 0.4900 Iter 13362, loss: 0.4638 Iter 13363, loss: 0.4910 Iter 13364, loss: 0.4618 Iter 13365, loss: 0.4927 Iter 13366, loss: 0.4825 Iter 13367, loss: 0.4596 Iter 13368, loss: 0.5089 Iter 13369, loss: 0.4604 Iter 13370, loss: 0.5163 Iter 13371, loss: 0.5232 Iter 13372, loss: 0.4947 Iter 13373, loss: 0.4881 Iter 13374, loss: 0.4606 Iter 13375, loss: 0.4864 Iter 13376, loss: 0.5072 Iter 13377, loss: 0.4930 Iter 13378, loss: 0.5172 Iter 13379, loss: 0.5315 Iter 13380, loss: 0.4875 Iter 13381, loss: 0.5423 Iter 13382, loss: 0.4320 Iter 13383, loss: 0.5048 Iter 13384, loss: 0.5406 Iter 13385, loss: 0.4879 Iter 13386, loss: 0.4736 Iter 13387, loss: 0.4742 Iter 13388, loss: 0.4721 Iter 13389, loss: 0.4864 Iter 13390, loss: 0.4992 Iter 13391, loss: 0.4987 Iter 13392, loss: 0.5180 Iter 13393, loss: 0.4736 Iter 13394, loss: 0.5014 Iter 13395, loss: 0.5144 Iter 13396, loss: 0.5456 Iter 13397, loss: 0.4726 Iter 13398, loss: 0.4704 Iter 13399, loss: 0.4921 Iter 13400, loss: 0.5227 Iter 13401, loss: 0.5101 Iter 13402, loss: 0.4646

Iter 13403, loss: 0.4917 Iter 13404, loss: 0.5060 Iter 13405, loss: 0.4608 Iter 13406, loss: 0.4577 Iter 13407, loss: 0.5338 Iter 13408, loss: 0.5187 Iter 13409, loss: 0.5260 Iter 13410, loss: 0.5056 Iter 13411, loss: 0.4904 Iter 13412, loss: 0.4712 Iter 13413, loss: 0.4721 Iter 13414, loss: 0.4766 Iter 13415, loss: 0.5445 Iter 13416, loss: 0.5120 Iter 13417, loss: 0.4939 Iter 13418, loss: 0.5046 Iter 13419, loss: 0.5019 Iter 13420, loss: 0.5030 Iter 13421, loss: 0.4635 Iter 13422, loss: 0.5160 Iter 13423, loss: 0.4888 Iter 13424, loss: 0.4826 Iter 13425, loss: 0.4640 Iter 13426, loss: 0.4969 Iter 13427, loss: 0.5122 Iter 13428, loss: 0.5032 Iter 13429, loss: 0.4838 Iter 13430, loss: 0.4789 Iter 13431, loss: 0.4779 Iter 13432, loss: 0.4759 Iter 13433, loss: 0.4451 Iter 13434, loss: 0.4826 Iter 13435, loss: 0.5127 Iter 13436, loss: 0.4854 Iter 13437, loss: 0.5135 Iter 13438, loss: 0.5435 Iter 13439, loss: 0.4893 Iter 13440, loss: 0.4909 Iter 13441, loss: 0.4635 Iter 13442, loss: 0.4664 Iter 13443, loss: 0.5007 Iter 13444, loss: 0.4866

Iter 13445, loss: 0.4851 Iter 13446, loss: 0.5519 Iter 13447, loss: 0.5316 Iter 13448, loss: 0.4790 Iter 13449, loss: 0.4405 Iter 13450, loss: 0.5055 Iter 13451, loss: 0.4881 Iter 13452, loss: 0.4618 Iter 13453, loss: 0.5377 Iter 13454, loss: 0.4765 Iter 13455, loss: 0.4802 Iter 13456, loss: 0.4938 Iter 13457, loss: 0.4766 Iter 13458, loss: 0.4859 Iter 13459, loss: 0.5030 Iter 13460, loss: 0.5050 Iter 13461, loss: 0.4962 Iter 13462, loss: 0.4658 Iter 13463, loss: 0.4394 Iter 13464, loss: 0.5099 Iter 13465, loss: 0.4818 Iter 13466, loss: 0.4557 Iter 13467, loss: 0.4976 Iter 13468, loss: 0.4865 Iter 13469, loss: 0.4970 Iter 13470, loss: 0.4946 Iter 13471, loss: 0.4920 Iter 13472, loss: 0.4724 Iter 13473, loss: 0.4890 Iter 13474, loss: 0.5417 Iter 13475, loss: 0.4976 Iter 13476, loss: 0.4754 Iter 13477, loss: 0.4670 Iter 13478, loss: 0.5122 Iter 13479, loss: 0.4928 Iter 13480, loss: 0.4883 Iter 13481, loss: 0.4659 Iter 13482, loss: 0.4761 Iter 13483, loss: 0.4980 Iter 13484, loss: 0.4926 Iter 13485, loss: 0.4977 Iter 13486, loss: 0.5194

Iter 13487, loss: 0.4933 Iter 13488, loss: 0.4880 Iter 13489, loss: 0.4793 Iter 13490, loss: 0.4855 Iter 13491, loss: 0.4891 Iter 13492, loss: 0.5269 Iter 13493, loss: 0.4857 Iter 13494, loss: 0.4754 Iter 13495, loss: 0.4662 Iter 13496, loss: 0.5124 Iter 13497, loss: 0.4713 Iter 13498, loss: 0.4895 Iter 13499, loss: 0.4568 Iter 13500, loss: 0.5213 Iter 13501, loss: 0.5245 Iter 13502, loss: 0.4952 Iter 13503, loss: 0.4980 Iter 13504, loss: 0.5054 Iter 13505, loss: 0.4793 Iter 13506, loss: 0.5089 Iter 13507, loss: 0.4940 Iter 13508, loss: 0.4867 Iter 13509, loss: 0.5033 Iter 13510, loss: 0.4720 Iter 13511, loss: 0.4990 Iter 13512, loss: 0.4629 Iter 13513, loss: 0.4839 Iter 13514, loss: 0.4733 Iter 13515, loss: 0.4813 Iter 13516, loss: 0.5253 Iter 13517, loss: 0.4543 Iter 13518, loss: 0.5218 Iter 13519, loss: 0.4767 Iter 13520, loss: 0.5225 Iter 13521, loss: 0.4626 Iter 13522, loss: 0.4521 Iter 13523, loss: 0.4769 Iter 13524, loss: 0.4868 Iter 13525, loss: 0.4795 Iter 13526, loss: 0.5136 Iter 13527, loss: 0.4985 Iter 13528, loss: 0.4966

Iter 13529, loss: 0.5003 Iter 13530, loss: 0.5153 Iter 13531, loss: 0.5556 Iter 13532, loss: 0.4858 Iter 13533, loss: 0.4796 Iter 13534, loss: 0.5074 Iter 13535, loss: 0.5011 Iter 13536, loss: 0.5018 Iter 13537, loss: 0.4630 Iter 13538, loss: 0.5300 Iter 13539, loss: 0.4869 Iter 13540, loss: 0.4903 Iter 13541, loss: 0.4613 Iter 13542, loss: 0.4620 Iter 13543, loss: 0.4694 Iter 13544, loss: 0.5080 Iter 13545, loss: 0.5037 Iter 13546, loss: 0.5066 Iter 13547, loss: 0.5127 Iter 13548, loss: 0.4486 Iter 13549, loss: 0.4808 Iter 13550, loss: 0.4785 Iter 13551, loss: 0.4740 Iter 13552, loss: 0.4749 Iter 13553, loss: 0.4944 Iter 13554, loss: 0.4944 Iter 13555, loss: 0.4868 Iter 13556, loss: 0.4831 Iter 13557, loss: 0.5062 Iter 13558, loss: 0.5085 Iter 13559, loss: 0.5140 Iter 13560, loss: 0.4599 Iter 13561, loss: 0.4901 Iter 13562, loss: 0.4736 Iter 13563, loss: 0.4853 Iter 13564, loss: 0.4841 Iter 13565, loss: 0.5301 Iter 13566, loss: 0.5205 Iter 13567, loss: 0.4973 Iter 13568, loss: 0.5053 Iter 13569, loss: 0.4898 Iter 13570, loss: 0.5108

Iter 13571, loss: 0.4871 Iter 13572, loss: 0.4821 Iter 13573, loss: 0.4906 Iter 13574, loss: 0.5018 Iter 13575, loss: 0.5169 Iter 13576, loss: 0.4811 Iter 13577, loss: 0.4979 Iter 13578, loss: 0.4634 Iter 13579, loss: 0.4788 Iter 13580, loss: 0.4674 Iter 13581, loss: 0.4581 Iter 13582, loss: 0.4463 Iter 13583, loss: 0.4716 Iter 13584, loss: 0.5501 Iter 13585, loss: 0.4844 Iter 13586, loss: 0.4655 Iter 13587, loss: 0.4912 Iter 13588, loss: 0.4891 Iter 13589, loss: 0.4706 Iter 13590, loss: 0.4562 Iter 13591, loss: 0.4631 Iter 13592, loss: 0.4953 Iter 13593, loss: 0.5053 Iter 13594, loss: 0.4620 Iter 13595, loss: 0.5138 Iter 13596, loss: 0.4788 Iter 13597, loss: 0.4917 Iter 13598, loss: 0.5414 Iter 13599, loss: 0.5135 Iter 13600, loss: 0.4973 Iter 13601, loss: 0.5069 Iter 13602, loss: 0.4631 Iter 13603, loss: 0.4903 Iter 13604, loss: 0.5109 Iter 13605, loss: 0.4646 Iter 13606, loss: 0.5259 Iter 13607, loss: 0.5200 Iter 13608, loss: 0.4846 Iter 13609, loss: 0.4679 Iter 13610, loss: 0.4740 Iter 13611, loss: 0.4675 Iter 13612, loss: 0.4673

Iter 13613, loss: 0.4795 Iter 13614, loss: 0.4866 Iter 13615, loss: 0.5116 Iter 13616, loss: 0.4769 Iter 13617, loss: 0.4850 Iter 13618, loss: 0.5339 Iter 13619, loss: 0.4518 Iter 13620, loss: 0.4419 Iter 13621, loss: 0.4907 Iter 13622, loss: 0.4901 Iter 13623, loss: 0.4893 Iter 13624, loss: 0.4742 Iter 13625, loss: 0.5290 Iter 13626, loss: 0.4835 Iter 13627, loss: 0.5107 Iter 13628, loss: 0.5099 Iter 13629, loss: 0.4978 Iter 13630, loss: 0.4750 Iter 13631, loss: 0.4841 Iter 13632, loss: 0.4806 Iter 13633, loss: 0.4826 Iter 13634, loss: 0.4938 Iter 13635, loss: 0.4946 Iter 13636, loss: 0.5048 Iter 13637, loss: 0.5256 Iter 13638, loss: 0.4627 Iter 13639, loss: 0.4839 Iter 13640, loss: 0.4489 Iter 13641, loss: 0.4612 Iter 13642, loss: 0.5050 Iter 13643, loss: 0.5138 Iter 13644, loss: 0.5128 Iter 13645, loss: 0.5019 Iter 13646, loss: 0.4880 Iter 13647, loss: 0.4834 Iter 13648, loss: 0.4372 Iter 13649, loss: 0.4749 Iter 13650, loss: 0.4938 Iter 13651, loss: 0.5191 Iter 13652, loss: 0.4751 Iter 13653, loss: 0.5125

Iter 13654, loss: 0.4784 Iter 13655, loss: 0.4929 Iter 13656, loss: 0.4466 Iter 13657, loss: 0.4439 Iter 13658, loss: 0.4983 Iter 13659, loss: 0.4607 Iter 13660, loss: 0.5131 Iter 13661, loss: 0.5190 Iter 13662, loss: 0.4801 Iter 13663, loss: 0.4802 Iter 13664, loss: 0.4729 Iter 13665, loss: 0.5256 Iter 13666, loss: 0.4943 Iter 13667, loss: 0.5092 Iter 13668, loss: 0.5077 Iter 13669, loss: 0.4932 Iter 13670, loss: 0.5073 Iter 13671, loss: 0.4665 Iter 13672, loss: 0.4610 Iter 13673, loss: 0.4498 Iter 13674, loss: 0.4883 Iter 13675, loss: 0.5313 Iter 13676, loss: 0.4844 Iter 13677, loss: 0.4983 Iter 13678, loss: 0.4883 Iter 13679, loss: 0.4970 Iter 13680, loss: 0.4490 Iter 13681, loss: 0.5124 Iter 13682, loss: 0.4854 Iter 13683, loss: 0.4547 Iter 13684, loss: 0.5129 Iter 13685, loss: 0.4759 Iter 13686, loss: 0.5038 Iter 13687, loss: 0.4852 Iter 13688, loss: 0.4926 Iter 13689, loss: 0.4588 Iter 13690, loss: 0.5408 Iter 13691, loss: 0.5236 Iter 13692, loss: 0.4854 Iter 13693, loss: 0.5263 Iter 13694, loss: 0.5082 Iter 13695, loss: 0.4901

Iter 13696, loss: 0.4735 Iter 13697, loss: 0.4745 Iter 13698, loss: 0.4879 Iter 13699, loss: 0.4829 Iter 13700, loss: 0.5065 Iter 13701, loss: 0.5294 Iter 13702, loss: 0.4406 Iter 13703, loss: 0.4728 Iter 13704, loss: 0.4467 Iter 13705, loss: 0.5042 Iter 13706, loss: 0.4902 Iter 13707, loss: 0.4695 Iter 13708, loss: 0.4801 Iter 13709, loss: 0.5159 Iter 13710, loss: 0.5054 Iter 13711, loss: 0.5197 Iter 13712, loss: 0.4910 Iter 13713, loss: 0.5019 Iter 13714, loss: 0.4619 Iter 13715, loss: 0.5059 Iter 13716, loss: 0.5135 Iter 13717, loss: 0.5278 Iter 13718, loss: 0.4922 Iter 13719, loss: 0.5313 Iter 13720, loss: 0.4922 Iter 13721, loss: 0.4905 Iter 13722, loss: 0.5412 Iter 13723, loss: 0.4854 Iter 13724, loss: 0.5007 Iter 13725, loss: 0.5085 Iter 13726, loss: 0.4982 Iter 13727, loss: 0.4804 Iter 13728, loss: 0.4698 Iter 13729, loss: 0.4716 Iter 13730, loss: 0.5291 Iter 13731, loss: 0.4893 Iter 13732, loss: 0.4464 Iter 13733, loss: 0.4852 Iter 13734, loss: 0.4824 Iter 13735, loss: 0.4769 Iter 13736, loss: 0.5311 Iter 13737, loss: 0.4775

Iter 13738, loss: 0.5057 Iter 13739, loss: 0.5255 Iter 13740, loss: 0.5029 Iter 13741, loss: 0.4668 Iter 13742, loss: 0.4818 Iter 13743, loss: 0.4922 Iter 13744, loss: 0.4928 Iter 13745, loss: 0.5193 Iter 13746, loss: 0.4654 Iter 13747, loss: 0.4984 Iter 13748, loss: 0.4639 Iter 13749, loss: 0.5099 Iter 13750, loss: 0.5170 Iter 13751, loss: 0.4791 Iter 13752, loss: 0.5003 Iter 13753, loss: 0.5074 Iter 13754, loss: 0.4898 Iter 13755, loss: 0.5017 Iter 13756, loss: 0.4959 Iter 13757, loss: 0.5124 Iter 13758, loss: 0.4465 Iter 13759, loss: 0.5039 Iter 13760, loss: 0.5229 Iter 13761, loss: 0.5170 Iter 13762, loss: 0.4481 Iter 13763, loss: 0.5042 Iter 13764, loss: 0.5165 Iter 13765, loss: 0.5278 Iter 13766, loss: 0.4679 Iter 13767, loss: 0.4677 Iter 13768, loss: 0.5000 Iter 13769, loss: 0.4723 Iter 13770, loss: 0.4691 Iter 13771, loss: 0.4585 Iter 13772, loss: 0.4688 Iter 13773, loss: 0.5088 Iter 13774, loss: 0.4543 Iter 13775, loss: 0.4500 Iter 13776, loss: 0.5437 Iter 13777, loss: 0.4925 Iter 13778, loss: 0.4998 Iter 13779, loss: 0.5058

Iter 13780, loss: 0.4686 Iter 13781, loss: 0.5049 Iter 13782, loss: 0.5058 Iter 13783, loss: 0.4641 Iter 13784, loss: 0.5235 Iter 13785, loss: 0.4696 Iter 13786, loss: 0.5206 Iter 13787, loss: 0.4962 Iter 13788, loss: 0.5501 Iter 13789, loss: 0.4646 Iter 13790, loss: 0.5115 Iter 13791, loss: 0.5099 Iter 13792, loss: 0.4945 Iter 13793, loss: 0.4924 Iter 13794, loss: 0.5178 Iter 13795, loss: 0.5053 Iter 13796, loss: 0.5106 Iter 13797, loss: 0.5258 Iter 13798, loss: 0.4819 Iter 13799, loss: 0.4841 Iter 13800, loss: 0.4696 Iter 13801, loss: 0.4691 Iter 13802, loss: 0.4939 Iter 13803, loss: 0.4766 Iter 13804, loss: 0.4954 Iter 13805, loss: 0.4675 Iter 13806, loss: 0.4636 Iter 13807, loss: 0.5131 Iter 13808, loss: 0.5054 Iter 13809, loss: 0.4650 Iter 13810, loss: 0.5114 Iter 13811, loss: 0.4715 Iter 13812, loss: 0.4691 Iter 13813, loss: 0.4865 Iter 13814, loss: 0.5137 Iter 13815, loss: 0.5258 Iter 13816, loss: 0.4999 Iter 13817, loss: 0.4703 Iter 13818, loss: 0.4780 Iter 13819, loss: 0.4794 Iter 13820, loss: 0.4611 Iter 13821, loss: 0.4641

Iter 13822, loss: 0.4994 Iter 13823, loss: 0.5533 Iter 13824, loss: 0.4739 Iter 13825, loss: 0.4775 Iter 13826, loss: 0.4884 Iter 13827, loss: 0.5097 Iter 13828, loss: 0.4736 Iter 13829, loss: 0.4787 Iter 13830, loss: 0.5309 Iter 13831, loss: 0.5080 Iter 13832, loss: 0.5034 Iter 13833, loss: 0.4456 Iter 13834, loss: 0.4645 Iter 13835, loss: 0.4504 Iter 13836, loss: 0.4660 Iter 13837, loss: 0.4513 Iter 13838, loss: 0.5289 Iter 13839, loss: 0.4642 Iter 13840, loss: 0.4756 Iter 13841, loss: 0.4965 Iter 13842, loss: 0.4945 Iter 13843, loss: 0.4696 Iter 13844, loss: 0.4790 Iter 13845, loss: 0.4945 Iter 13846, loss: 0.5050 Iter 13847, loss: 0.5139 Iter 13848, loss: 0.4905 Iter 13849, loss: 0.5455 Iter 13850, loss: 0.5006 Iter 13851, loss: 0.4448 Iter 13852, loss: 0.4983 Iter 13853, loss: 0.4972 Iter 13854, loss: 0.4581 Iter 13855, loss: 0.4698 Iter 13856, loss: 0.4621 Iter 13857, loss: 0.5001 Iter 13858, loss: 0.5000 Iter 13859, loss: 0.4950 Iter 13860, loss: 0.4981 Iter 13861, loss: 0.5045 Iter 13862, loss: 0.4777

Iter 13863, loss: 0.4889 Iter 13864, loss: 0.4708 Iter 13865, loss: 0.4651 Iter 13866, loss: 0.5101 Iter 13867, loss: 0.5051 Iter 13868, loss: 0.4580 Iter 13869, loss: 0.5152 Iter 13870, loss: 0.5054 Iter 13871, loss: 0.4991 Iter 13872, loss: 0.4971 Iter 13873, loss: 0.4617 Iter 13874, loss: 0.5224 Iter 13875, loss: 0.4492 Iter 13876, loss: 0.4741 Iter 13877, loss: 0.5161 Iter 13878, loss: 0.4946 Iter 13879, loss: 0.5211 Iter 13880, loss: 0.5262 Iter 13881, loss: 0.5079 Iter 13882, loss: 0.4607 Iter 13883, loss: 0.5071 Iter 13884, loss: 0.4721 Iter 13885, loss: 0.4432 Iter 13886, loss: 0.5285 Iter 13887, loss: 0.5028 Iter 13888, loss: 0.4837 Iter 13889, loss: 0.5021 Iter 13890, loss: 0.4664 Iter 13891, loss: 0.4797 Iter 13892, loss: 0.4742 Iter 13893, loss: 0.5242 Iter 13894, loss: 0.4811 Iter 13895, loss: 0.5352 Iter 13896, loss: 0.5366 Iter 13897, loss: 0.5045 Iter 13898, loss: 0.4918 Iter 13899, loss: 0.5201 Iter 13900, loss: 0.5052 Iter 13901, loss: 0.4440 Iter 13902, loss: 0.4772 Iter 13903, loss: 0.4672 Iter 13904, loss: 0.4925

Iter 13905, loss: 0.4701 Iter 13906, loss: 0.4853 Iter 13907, loss: 0.4699 Iter 13908, loss: 0.4519 Iter 13909, loss: 0.5246 Iter 13910, loss: 0.4839 Iter 13911, loss: 0.4908 Iter 13912, loss: 0.5434 Iter 13913, loss: 0.4758 Iter 13914, loss: 0.4613 Iter 13915, loss: 0.5072 Iter 13916, loss: 0.5053 Iter 13917, loss: 0.4999 Iter 13918, loss: 0.4516 Iter 13919, loss: 0.5066 Iter 13920, loss: 0.4873 Iter 13921, loss: 0.4752 Iter 13922, loss: 0.5201 Iter 13923, loss: 0.5457 Iter 13924, loss: 0.4948 Iter 13925, loss: 0.4827 Iter 13926, loss: 0.4728 Iter 13927, loss: 0.4766 Iter 13928, loss: 0.4888 Iter 13929, loss: 0.4689 Iter 13930, loss: 0.4871 Iter 13931, loss: 0.4897 Iter 13932, loss: 0.4779 Iter 13933, loss: 0.4917 Iter 13934, loss: 0.4758 Iter 13935, loss: 0.4954 Iter 13936, loss: 0.5040 Iter 13937, loss: 0.5296 Iter 13938, loss: 0.4917 Iter 13939, loss: 0.4792 Iter 13940, loss: 0.4870 Iter 13941, loss: 0.4813 Iter 13942, loss: 0.5049 Iter 13943, loss: 0.4737 Iter 13944, loss: 0.4981 Iter 13945, loss: 0.4622 Iter 13946, loss: 0.4679

Iter 13947, loss: 0.4902 Iter 13948, loss: 0.4209 Iter 13949, loss: 0.4667 Iter 13950, loss: 0.5083 Iter 13951, loss: 0.4253 Iter 13952, loss: 0.4618 Iter 13953, loss: 0.4855 Iter 13954, loss: 0.5238 Iter 13955, loss: 0.4899 Iter 13956, loss: 0.5105 Iter 13957, loss: 0.4800 Iter 13958, loss: 0.4595 Iter 13959, loss: 0.5274 Iter 13960, loss: 0.4536 Iter 13961, loss: 0.4753 Iter 13962, loss: 0.5051 Iter 13963, loss: 0.4827 Iter 13964, loss: 0.4925 Iter 13965, loss: 0.4958 Iter 13966, loss: 0.4851 Iter 13967, loss: 0.4571 Iter 13968, loss: 0.4917 Iter 13969, loss: 0.4880 Iter 13970, loss: 0.4592 Iter 13971, loss: 0.4550 Iter 13972, loss: 0.5117 Iter 13973, loss: 0.4524 Iter 13974, loss: 0.5119 Iter 13975, loss: 0.4950 Iter 13976, loss: 0.4334 Iter 13977, loss: 0.4625 Iter 13978, loss: 0.4616 Iter 13979, loss: 0.4833 Iter 13980, loss: 0.4645 Iter 13981, loss: 0.5018 Iter 13982, loss: 0.4843 Iter 13983, loss: 0.5033 Iter 13984, loss: 0.4853 Iter 13985, loss: 0.4974 Iter 13986, loss: 0.4864 Iter 13987, loss: 0.4928 Iter 13988, loss: 0.5068

Iter 13989, loss: 0.4932 Iter 13990, loss: 0.4797 Iter 13991, loss: 0.4996 Iter 13992, loss: 0.4784 Iter 13993, loss: 0.4848 Iter 13994, loss: 0.4696 Iter 13995, loss: 0.4494 Iter 13996, loss: 0.5060 Iter 13997, loss: 0.5175 Iter 13998, loss: 0.4755 Iter 13999, loss: 0.4702 Iter 14000, loss: 0.5225 Iter 14001, loss: 0.4958 Iter 14002, loss: 0.4895 Iter 14003, loss: 0.4717 Iter 14004, loss: 0.5120 Iter 14005, loss: 0.4742 Iter 14006, loss: 0.4678 Iter 14007, loss: 0.5203 Iter 14008, loss: 0.5047 Iter 14009, loss: 0.4848 Iter 14010, loss: 0.4884 Iter 14011, loss: 0.4925 Iter 14012, loss: 0.4311 Iter 14013, loss: 0.4856 Iter 14014, loss: 0.4892 Iter 14015, loss: 0.5098 Iter 14016, loss: 0.5366 Iter 14017, loss: 0.4663 Iter 14018, loss: 0.4788 Iter 14019, loss: 0.4793 Iter 14020, loss: 0.4316 Iter 14021, loss: 0.4802 Iter 14022, loss: 0.4780 Iter 14023, loss: 0.4924 Iter 14024, loss: 0.4994 Iter 14025, loss: 0.5084 Iter 14026, loss: 0.4970 Iter 14027, loss: 0.5112 Iter 14028, loss: 0.4535 Iter 14029, loss: 0.4931 Iter 14030, loss: 0.5089

Iter 14031, loss: 0.5101 Iter 14032, loss: 0.5084 Iter 14033, loss: 0.4661 Iter 14034, loss: 0.4873 Iter 14035, loss: 0.4996 Iter 14036, loss: 0.4889 Iter 14037, loss: 0.4819 Iter 14038, loss: 0.4582 Iter 14039, loss: 0.4864 Iter 14040, loss: 0.4892 Iter 14041, loss: 0.5355 Iter 14042, loss: 0.4932 Iter 14043, loss: 0.4533 Iter 14044, loss: 0.4485 Iter 14045, loss: 0.4983 Iter 14046, loss: 0.4985 Iter 14047, loss: 0.4590 Iter 14048, loss: 0.4648 Iter 14049, loss: 0.4864 Iter 14050, loss: 0.4752 Iter 14051, loss: 0.5074 Iter 14052, loss: 0.4656 Iter 14053, loss: 0.4703 Iter 14054, loss: 0.4824 Iter 14055, loss: 0.4802 Iter 14056, loss: 0.4701 Iter 14057, loss: 0.4814 Iter 14058, loss: 0.4964 Iter 14059, loss: 0.5038 Iter 14060, loss: 0.4665 Iter 14061, loss: 0.4814 Iter 14062, loss: 0.4665 Iter 14063, loss: 0.4609 Iter 14064, loss: 0.4631 Iter 14065, loss: 0.5162 Iter 14066, loss: 0.4995 Iter 14067, loss: 0.4590 Iter 14068, loss: 0.4954 Iter 14069, loss: 0.5007 Iter 14070, loss: 0.5004 Iter 14071, loss: 0.5074 Iter 14072, loss: 0.4740

Iter 14073, loss: 0.4726 Iter 14074, loss: 0.4433 Iter 14075, loss: 0.4967 Iter 14076, loss: 0.4946 Iter 14077, loss: 0.4489 Iter 14078, loss: 0.5030 Iter 14079, loss: 0.4729 Iter 14080, loss: 0.4750 Iter 14081, loss: 0.4820 Iter 14082, loss: 0.5176 Iter 14083, loss: 0.4565 Iter 14084, loss: 0.4719 Iter 14085, loss: 0.4867 Iter 14086, loss: 0.5072 Iter 14087, loss: 0.5230 Iter 14088, loss: 0.4878 Iter 14089, loss: 0.5143 Iter 14090, loss: 0.4977 Iter 14091, loss: 0.4951 Iter 14092, loss: 0.4843 Iter 14093, loss: 0.5095 Iter 14094, loss: 0.4498 Iter 14095, loss: 0.4758 Iter 14096, loss: 0.5179 Iter 14097, loss: 0.4897 Iter 14098, loss: 0.4305 Iter 14099, loss: 0.4515 Iter 14100, loss: 0.4872 Iter 14101, loss: 0.4868 Iter 14102, loss: 0.4849 Iter 14103, loss: 0.4990 Iter 14104, loss: 0.5551 Iter 14105, loss: 0.5022 Iter 14106, loss: 0.4913 Iter 14107, loss: 0.4765 Iter 14108, loss: 0.5033 Iter 14109, loss: 0.4871 Iter 14110, loss: 0.4628 Iter 14111, loss: 0.4453 Iter 14112, loss: 0.4881 Iter 14113, loss: 0.5100 Iter 14114, loss: 0.4689

Iter 14115, loss: 0.4825 Iter 14116, loss: 0.4999 Iter 14117, loss: 0.4699 Iter 14118, loss: 0.4567 Iter 14119, loss: 0.4854 Iter 14120, loss: 0.4763 Iter 14121, loss: 0.4688 Iter 14122, loss: 0.4642 Iter 14123, loss: 0.4617 Iter 14124, loss: 0.4615 Iter 14125, loss: 0.4786 Iter 14126, loss: 0.4794 Iter 14127, loss: 0.5144 Iter 14128, loss: 0.4842 Iter 14129, loss: 0.4791 Iter 14130, loss: 0.5206 Iter 14131, loss: 0.4945 Iter 14132, loss: 0.4897 Iter 14133, loss: 0.5445 Iter 14134, loss: 0.4638 Iter 14135, loss: 0.5229 Iter 14136, loss: 0.4886 Iter 14137, loss: 0.5027 Iter 14138, loss: 0.5008 Iter 14139, loss: 0.5317 Iter 14140, loss: 0.4607 Iter 14141, loss: 0.5081 Iter 14142, loss: 0.5051 Iter 14143, loss: 0.4655 Iter 14144, loss: 0.4970 Iter 14145, loss: 0.5083 Iter 14146, loss: 0.5148 Iter 14147, loss: 0.4968 Iter 14148, loss: 0.4973 Iter 14149, loss: 0.5142 Iter 14150, loss: 0.4933 Iter 14151, loss: 0.5011 Iter 14152, loss: 0.4787 Iter 14153, loss: 0.4797 Iter 14154, loss: 0.4433 Iter 14155, loss: 0.4939 Iter 14156, loss: 0.5244

Iter 14157, loss: 0.5264 Iter 14158, loss: 0.5032 Iter 14159, loss: 0.4612 Iter 14160, loss: 0.5317 Iter 14161, loss: 0.5284 Iter 14162, loss: 0.5020 Iter 14163, loss: 0.5066 Iter 14164, loss: 0.4658 Iter 14165, loss: 0.5081 Iter 14166, loss: 0.4962 Iter 14167, loss: 0.4961 Iter 14168, loss: 0.5224 Iter 14169, loss: 0.4675 Iter 14170, loss: 0.4705 Iter 14171, loss: 0.4382 Iter 14172, loss: 0.4670 Iter 14173, loss: 0.4941 Iter 14174, loss: 0.4764 Iter 14175, loss: 0.4856 Iter 14176, loss: 0.4901 Iter 14177, loss: 0.4724 Iter 14178, loss: 0.4891 Iter 14179, loss: 0.4979 Iter 14180, loss: 0.4640 Iter 14181, loss: 0.4544 Iter 14182, loss: 0.4870 Iter 14183, loss: 0.4928 Iter 14184, loss: 0.4768 Iter 14185, loss: 0.4612 Iter 14186, loss: 0.4829 Iter 14187, loss: 0.5280 Iter 14188, loss: 0.4665 Iter 14189, loss: 0.4569 Iter 14190, loss: 0.4997 Iter 14191, loss: 0.4485 Iter 14192, loss: 0.4672 Iter 14193, loss: 0.4820 Iter 14194, loss: 0.4691 Iter 14195, loss: 0.4834 Iter 14196, loss: 0.5265 Iter 14197, loss: 0.4342 Iter 14198, loss: 0.4604

Iter 14199, loss: 0.5025 Iter 14200, loss: 0.4617 Iter 14201, loss: 0.5005 Iter 14202, loss: 0.4948 Iter 14203, loss: 0.5322 Iter 14204, loss: 0.4678 Iter 14205, loss: 0.4812 Iter 14206, loss: 0.5052 Iter 14207, loss: 0.4533 Iter 14208, loss: 0.5169 Iter 14209, loss: 0.4866 Iter 14210, loss: 0.5165 Iter 14211, loss: 0.4620 Iter 14212, loss: 0.4883 Iter 14213, loss: 0.5171 Iter 14214, loss: 0.5130 Iter 14215, loss: 0.5130 Iter 14216, loss: 0.5042 Iter 14217, loss: 0.4889 Iter 14218, loss: 0.4965 Iter 14219, loss: 0.4765 Iter 14220, loss: 0.4684 Iter 14221, loss: 0.5104 Iter 14222, loss: 0.4891 Iter 14223, loss: 0.5017 Iter 14224, loss: 0.4632 Iter 14225, loss: 0.4946 Iter 14226, loss: 0.4991 Iter 14227, loss: 0.4921 Iter 14228, loss: 0.4810 Iter 14229, loss: 0.4984 Iter 14230, loss: 0.4854 Iter 14231, loss: 0.4813 Iter 14232, loss: 0.4999 Iter 14233, loss: 0.4851 Iter 14234, loss: 0.4539 Iter 14235, loss: 0.4570 Iter 14236, loss: 0.4700 Iter 14237, loss: 0.5242 Iter 14238, loss: 0.4559 Iter 14239, loss: 0.4845 Iter 14240, loss: 0.4955

Iter 14241, loss: 0.4965 Iter 14242, loss: 0.5000 Iter 14243, loss: 0.4758 Iter 14244, loss: 0.4574 Iter 14245, loss: 0.4711 Iter 14246, loss: 0.4434 Iter 14247, loss: 0.4927 Iter 14248, loss: 0.5326 Iter 14249, loss: 0.4771 Iter 14250, loss: 0.4922 Iter 14251, loss: 0.4849 Iter 14252, loss: 0.4801 Iter 14253, loss: 0.4729 Iter 14254, loss: 0.5065 Iter 14255, loss: 0.4889 Iter 14256, loss: 0.4948 Iter 14257, loss: 0.4582 Iter 14258, loss: 0.5228 Iter 14259, loss: 0.5176 Iter 14260, loss: 0.4732 Iter 14261, loss: 0.4735 Iter 14262, loss: 0.4646 Iter 14263, loss: 0.4894 Iter 14264, loss: 0.4713 Iter 14265, loss: 0.4707 Iter 14266, loss: 0.5259 Iter 14267, loss: 0.4898 Iter 14268, loss: 0.4721 Iter 14269, loss: 0.4357 Iter 14270, loss: 0.4941 Iter 14271, loss: 0.4938 Iter 14272, loss: 0.4905 Iter 14273, loss: 0.4834 Iter 14274, loss: 0.5085 Iter 14275, loss: 0.4515 Iter 14276, loss: 0.4927 Iter 14277, loss: 0.5134 Iter 14278, loss: 0.4936 Iter 14279, loss: 0.5016 Iter 14280, loss: 0.4588 Iter 14281, loss: 0.4705 Iter 14282, loss: 0.4690

Iter 14283, loss: 0.4585 Iter 14284, loss: 0.4688 Iter 14285, loss: 0.5304 Iter 14286, loss: 0.4858 Iter 14287, loss: 0.4791 Iter 14288, loss: 0.4909 Iter 14289, loss: 0.4509 Iter 14290, loss: 0.5294 Iter 14291, loss: 0.4818 Iter 14292, loss: 0.5159 Iter 14293, loss: 0.4715 Iter 14294, loss: 0.4997 Iter 14295, loss: 0.4814 Iter 14296, loss: 0.4717 Iter 14297, loss: 0.4567 Iter 14298, loss: 0.4726 Iter 14299, loss: 0.4815 Iter 14300, loss: 0.4818 Iter 14301, loss: 0.4559 Iter 14302, loss: 0.5218 Iter 14303, loss: 0.4849 Iter 14304, loss: 0.4928 Iter 14305, loss: 0.5259 Iter 14306, loss: 0.4745 Iter 14307, loss: 0.5088 Iter 14308, loss: 0.5033 Iter 14309, loss: 0.4982 Iter 14310, loss: 0.4885 Iter 14311, loss: 0.5155 Iter 14312, loss: 0.4713 Iter 14313, loss: 0.4709 Iter 14314, loss: 0.4928 Iter 14315, loss: 0.5080 Iter 14316, loss: 0.5054 Iter 14317, loss: 0.4662 Iter 14318, loss: 0.5047 Iter 14319, loss: 0.5618 Iter 14320, loss: 0.4822 Iter 14321, loss: 0.4835 Iter 14322, loss: 0.5185 Iter 14323, loss: 0.5373 Iter 14324, loss: 0.4970

Iter 14325, loss: 0.4970 Iter 14326, loss: 0.4603 Iter 14327, loss: 0.5209 Iter 14328, loss: 0.4654 Iter 14329, loss: 0.4868 Iter 14330, loss: 0.4752 Iter 14331, loss: 0.4955 Iter 14332, loss: 0.4853 Iter 14333, loss: 0.5285 Iter 14334, loss: 0.4992 Iter 14335, loss: 0.5046 Iter 14336, loss: 0.4891 Iter 14337, loss: 0.5009 Iter 14338, loss: 0.4788 Iter 14339, loss: 0.4776 Iter 14340, loss: 0.5153 Iter 14341, loss: 0.5125 Iter 14342, loss: 0.5346 Iter 14343, loss: 0.4560 Iter 14344, loss: 0.5058 Iter 14345, loss: 0.4726 Iter 14346, loss: 0.4784 Iter 14347, loss: 0.4877 Iter 14348, loss: 0.4606 Iter 14349, loss: 0.4635 Iter 14350, loss: 0.4785 Iter 14351, loss: 0.4979 Iter 14352, loss: 0.5003 Iter 14353, loss: 0.5226 Iter 14354, loss: 0.4900 Iter 14355, loss: 0.4870 Iter 14356, loss: 0.4662 Iter 14357, loss: 0.4699 Iter 14358, loss: 0.5163 Iter 14359, loss: 0.5052 Iter 14360, loss: 0.4643 Iter 14361, loss: 0.4613 Iter 14362, loss: 0.5225 Iter 14363, loss: 0.4953 Iter 14364, loss: 0.5617 Iter 14365, loss: 0.5039 Iter 14366, loss: 0.4655

Iter 14367, loss: 0.4862 Iter 14368, loss: 0.4981 Iter 14369, loss: 0.4421 Iter 14370, loss: 0.5044 Iter 14371, loss: 0.4936 Iter 14372, loss: 0.4612 Iter 14373, loss: 0.4885 Iter 14374, loss: 0.4727 Iter 14375, loss: 0.4471 Iter 14376, loss: 0.4565 Iter 14377, loss: 0.4962 Iter 14378, loss: 0.4809 Iter 14379, loss: 0.4863 Iter 14380, loss: 0.4937 Iter 14381, loss: 0.5112 Iter 14382, loss: 0.4883 Iter 14383, loss: 0.4859 Iter 14384, loss: 0.4586 Iter 14385, loss: 0.4797 Iter 14386, loss: 0.5063 Iter 14387, loss: 0.5212 Iter 14388, loss: 0.5102 Iter 14389, loss: 0.4562 Iter 14390, loss: 0.4823 Iter 14391, loss: 0.4934 Iter 14392, loss: 0.4637 Iter 14393, loss: 0.4832 Iter 14394, loss: 0.5188 Iter 14395, loss: 0.5137 Iter 14396, loss: 0.5096 Iter 14397, loss: 0.4605 Iter 14398, loss: 0.4842 Iter 14399, loss: 0.4988 Iter 14400, loss: 0.4781 Iter 14401, loss: 0.5211 Iter 14402, loss: 0.4746 Iter 14403, loss: 0.5134 Iter 14404, loss: 0.4621 Iter 14405, loss: 0.4735 Iter 14406, loss: 0.4796 Iter 14407, loss: 0.4913 Iter 14408, loss: 0.4864

Iter 14409, loss: 0.5041 Iter 14410, loss: 0.4805 Iter 14411, loss: 0.4836 Iter 14412, loss: 0.4836 Iter 14413, loss: 0.4894 Iter 14414, loss: 0.4877 Iter 14415, loss: 0.5045 Iter 14416, loss: 0.4825 Iter 14417, loss: 0.4647 Iter 14418, loss: 0.4779 Iter 14419, loss: 0.5158 Iter 14420, loss: 0.4968 Iter 14421, loss: 0.4871 Iter 14422, loss: 0.4977 Iter 14423, loss: 0.5053 Iter 14424, loss: 0.5111 Iter 14425, loss: 0.5014 Iter 14426, loss: 0.5247 Iter 14427, loss: 0.5198 Iter 14428, loss: 0.5204 Iter 14429, loss: 0.4849 Iter 14430, loss: 0.4534 Iter 14431, loss: 0.5160 Iter 14432, loss: 0.4534 Iter 14433, loss: 0.5030 Iter 14434, loss: 0.5007 Iter 14435, loss: 0.5017 Iter 14436, loss: 0.5071 Iter 14437, loss: 0.4641 Iter 14438, loss: 0.5005 Iter 14439, loss: 0.4900 Iter 14440, loss: 0.4837 Iter 14441, loss: 0.5142 Iter 14442, loss: 0.4617 Iter 14443, loss: 0.5103 Iter 14444, loss: 0.5043 Iter 14445, loss: 0.4725 Iter 14446, loss: 0.4520 Iter 14447, loss: 0.4900 Iter 14448, loss: 0.4912 Iter 14449, loss: 0.5123 Iter 14450, loss: 0.4519

Iter 14451, loss: 0.4973 Iter 14452, loss: 0.4664 Iter 14453, loss: 0.4799 Iter 14454, loss: 0.4712 Iter 14455, loss: 0.4667 Iter 14456, loss: 0.4839 Iter 14457, loss: 0.4928 Iter 14458, loss: 0.4964 Iter 14459, loss: 0.4993 Iter 14460, loss: 0.5141 Iter 14461, loss: 0.4828 Iter 14462, loss: 0.4770 Iter 14463, loss: 0.4883 Iter 14464, loss: 0.4702 Iter 14465, loss: 0.5116 Iter 14466, loss: 0.4606 Iter 14467, loss: 0.4540 Iter 14468, loss: 0.5325 Iter 14469, loss: 0.5089 Iter 14470, loss: 0.4578 Iter 14471, loss: 0.4917 Iter 14472, loss: 0.5050 Iter 14473, loss: 0.5113 Iter 14474, loss: 0.5369 Iter 14475, loss: 0.4546 Iter 14476, loss: 0.5162 Iter 14477, loss: 0.4713 Iter 14478, loss: 0.4888 Iter 14479, loss: 0.4867 Iter 14480, loss: 0.4617 Iter 14481, loss: 0.4978 Iter 14482, loss: 0.5025 Iter 14483, loss: 0.4983 Iter 14484, loss: 0.4779 Iter 14485, loss: 0.5110 Iter 14486, loss: 0.4796 Iter 14487, loss: 0.4256 Iter 14488, loss: 0.5354 Iter 14489, loss: 0.4924 Iter 14490, loss: 0.4925 Iter 14491, loss: 0.4789 Iter 14492, loss: 0.5119

Iter 14493, loss: 0.4851 Iter 14494, loss: 0.4852 Iter 14495, loss: 0.5092 Iter 14496, loss: 0.5359 Iter 14497, loss: 0.4634 Iter 14498, loss: 0.4946 Iter 14499, loss: 0.4660 Iter 14500, loss: 0.5078 Iter 14501, loss: 0.4681 Iter 14502, loss: 0.4695 Iter 14503, loss: 0.4889 Iter 14504, loss: 0.4839 Iter 14505, loss: 0.5095 Iter 14506, loss: 0.4530 Iter 14507, loss: 0.4614 Iter 14508, loss: 0.4754 Iter 14509, loss: 0.4624 Iter 14510, loss: 0.4926 Iter 14511, loss: 0.4895 Iter 14512, loss: 0.5024 Iter 14513, loss: 0.4925 Iter 14514, loss: 0.5184 Iter 14515, loss: 0.4877 Iter 14516, loss: 0.4691 Iter 14517, loss: 0.5193 Iter 14518, loss: 0.4786 Iter 14519, loss: 0.5013 Iter 14520, loss: 0.4968 Iter 14521, loss: 0.4933 Iter 14522, loss: 0.4959 Iter 14523, loss: 0.4864 Iter 14524, loss: 0.4863 Iter 14525, loss: 0.4412 Iter 14526, loss: 0.4839 Iter 14527, loss: 0.4850 Iter 14528, loss: 0.4761 Iter 14529, loss: 0.5162 Iter 14530, loss: 0.4987 Iter 14531, loss: 0.4882 Iter 14532, loss: 0.5019 Iter 14533, loss: 0.4850 Iter 14534, loss: 0.4526

Iter 14535, loss: 0.4727 Iter 14536, loss: 0.4826 Iter 14537, loss: 0.4789 Iter 14538, loss: 0.4674 Iter 14539, loss: 0.4418 Iter 14540, loss: 0.4728 Iter 14541, loss: 0.4578 Iter 14542, loss: 0.4879 Iter 14543, loss: 0.4872 Iter 14544, loss: 0.4636 Iter 14545, loss: 0.4748 Iter 14546, loss: 0.4744 Iter 14547, loss: 0.4675 Iter 14548, loss: 0.4621 Iter 14549, loss: 0.5083 Iter 14550, loss: 0.4665 Iter 14551, loss: 0.5098 Iter 14552, loss: 0.4964 Iter 14553, loss: 0.4691 Iter 14554, loss: 0.4842 Iter 14555, loss: 0.5123 Iter 14556, loss: 0.5149 Iter 14557, loss: 0.4899 Iter 14558, loss: 0.5240 Iter 14559, loss: 0.4485 Iter 14560, loss: 0.4576 Iter 14561, loss: 0.4870 Iter 14562, loss: 0.4566 Iter 14563, loss: 0.5029 Iter 14564, loss: 0.4696 Iter 14565, loss: 0.4860 Iter 14566, loss: 0.4753 Iter 14567, loss: 0.5138 Iter 14568, loss: 0.4763 Iter 14569, loss: 0.5080 Iter 14570, loss: 0.4499 Iter 14571, loss: 0.4990 Iter 14572, loss: 0.4956 Iter 14573, loss: 0.4872 Iter 14574, loss: 0.4985 Iter 14575, loss: 0.4844 Iter 14576, loss: 0.5185

Iter 14577, loss: 0.4550 Iter 14578, loss: 0.4882 Iter 14579, loss: 0.4835 Iter 14580, loss: 0.5004 Iter 14581, loss: 0.4946 Iter 14582, loss: 0.5215 Iter 14583, loss: 0.4767 Iter 14584, loss: 0.5163 Iter 14585, loss: 0.5038 Iter 14586, loss: 0.4723 Iter 14587, loss: 0.4811 Iter 14588, loss: 0.4908 Iter 14589, loss: 0.4456 Iter 14590, loss: 0.5096 Iter 14591, loss: 0.4646 Iter 14592, loss: 0.4852 Iter 14593, loss: 0.4931 Iter 14594, loss: 0.5370 Iter 14595, loss: 0.4918 Iter 14596, loss: 0.4726 Iter 14597, loss: 0.4791 Iter 14598, loss: 0.4892 Iter 14599, loss: 0.4805 Iter 14600, loss: 0.5046 Iter 14601, loss: 0.4709 Iter 14602, loss: 0.4905 Iter 14603, loss: 0.5030 Iter 14604, loss: 0.4814 Iter 14605, loss: 0.4643 Iter 14606, loss: 0.4728 Iter 14607, loss: 0.4531 Iter 14608, loss: 0.4548 Iter 14609, loss: 0.4566 Iter 14610, loss: 0.4911 Iter 14611, loss: 0.5240 Iter 14612, loss: 0.5280 Iter 14613, loss: 0.4916 Iter 14614, loss: 0.4913 Iter 14615, loss: 0.4991 Iter 14616, loss: 0.5085 Iter 14617, loss: 0.4986 Iter 14618, loss: 0.4880

Iter 14619, loss: 0.4767 Iter 14620, loss: 0.4659 Iter 14621, loss: 0.4748 Iter 14622, loss: 0.4641 Iter 14623, loss: 0.4356 Iter 14624, loss: 0.4895 Iter 14625, loss: 0.4903 Iter 14626, loss: 0.5007 Iter 14627, loss: 0.4998 Iter 14628, loss: 0.4704 Iter 14629, loss: 0.4915 Iter 14630, loss: 0.4573 Iter 14631, loss: 0.4470 Iter 14632, loss: 0.4969 Iter 14633, loss: 0.4696 Iter 14634, loss: 0.4562 Iter 14635, loss: 0.4699 Iter 14636, loss: 0.4600 Iter 14637, loss: 0.4913 Iter 14638, loss: 0.4990 Iter 14639, loss: 0.5124 Iter 14640, loss: 0.4817 Iter 14641, loss: 0.4904 Iter 14642, loss: 0.5033 Iter 14643, loss: 0.4764 Iter 14644, loss: 0.4968 Iter 14645, loss: 0.5161 Iter 14646, loss: 0.4803 Iter 14647, loss: 0.5221 Iter 14648, loss: 0.4851 Iter 14649, loss: 0.4669 Iter 14650, loss: 0.4564 Iter 14651, loss: 0.4973 Iter 14652, loss: 0.5011 Iter 14653, loss: 0.5026 Iter 14654, loss: 0.4813 Iter 14655, loss: 0.4902 Iter 14656, loss: 0.4646 Iter 14657, loss: 0.5076 Iter 14658, loss: 0.4923 Iter 14659, loss: 0.4748 Iter 14660, loss: 0.4825

Iter 14661, loss: 0.5194 Iter 14662, loss: 0.4593 Iter 14663, loss: 0.4870 Iter 14664, loss: 0.4529 Iter 14665, loss: 0.5014 Iter 14666, loss: 0.4999 Iter 14667, loss: 0.4797 Iter 14668, loss: 0.4741 Iter 14669, loss: 0.5076 Iter 14670, loss: 0.5300 Iter 14671, loss: 0.4858 Iter 14672, loss: 0.4580 Iter 14673, loss: 0.4756 Iter 14674, loss: 0.4768 Iter 14675, loss: 0.4807 Iter 14676, loss: 0.4670 Iter 14677, loss: 0.4861 Iter 14678, loss: 0.4705 Iter 14679, loss: 0.4508 Iter 14680, loss: 0.4443 Iter 14681, loss: 0.5023 Iter 14682, loss: 0.4699 Iter 14683, loss: 0.4764 Iter 14684, loss: 0.4826 Iter 14685, loss: 0.4826 Iter 14686, loss: 0.5562 Iter 14687, loss: 0.5021 Iter 14688, loss: 0.4986 Iter 14689, loss: 0.5344 Iter 14690, loss: 0.4980 Iter 14691, loss: 0.4647 Iter 14692, loss: 0.4819 Iter 14693, loss: 0.4510 Iter 14694, loss: 0.5149 Iter 14695, loss: 0.4869 Iter 14696, loss: 0.4602 Iter 14697, loss: 0.5105 Iter 14698, loss: 0.4994 Iter 14699, loss: 0.5226 Iter 14700, loss: 0.5073 Iter 14701, loss: 0.4794 Iter 14702, loss: 0.5204

Iter 14703, loss: 0.4986 Iter 14704, loss: 0.5002 Iter 14705, loss: 0.4649 Iter 14706, loss: 0.4659 Iter 14707, loss: 0.4789 Iter 14708, loss: 0.4948 Iter 14709, loss: 0.4932 Iter 14710, loss: 0.4448 Iter 14711, loss: 0.4910 Iter 14712, loss: 0.4382 Iter 14713, loss: 0.5456 Iter 14714, loss: 0.4892 Iter 14715, loss: 0.4606 Iter 14716, loss: 0.4821 Iter 14717, loss: 0.4869 Iter 14718, loss: 0.4997 Iter 14719, loss: 0.4884 Iter 14720, loss: 0.5407 Iter 14721, loss: 0.4838 Iter 14722, loss: 0.4826 Iter 14723, loss: 0.4739 Iter 14724, loss: 0.4749 Iter 14725, loss: 0.5044 Iter 14726, loss: 0.4818 Iter 14727, loss: 0.5625 Iter 14728, loss: 0.5101 Iter 14729, loss: 0.4633 Iter 14730, loss: 0.4526 Iter 14731, loss: 0.4849 Iter 14732, loss: 0.5144 Iter 14733, loss: 0.4952 Iter 14734, loss: 0.4792 Iter 14735, loss: 0.4941 Iter 14736, loss: 0.4766 Iter 14737, loss: 0.4943 Iter 14738, loss: 0.4921 Iter 14739, loss: 0.4852 Iter 14740, loss: 0.4982 Iter 14741, loss: 0.4648 Iter 14742, loss: 0.4960 Iter 14743, loss: 0.4819 Iter 14744, loss: 0.4694

Iter 14745, loss: 0.4742 Iter 14746, loss: 0.4554 Iter 14747, loss: 0.5027 Iter 14748, loss: 0.5310 Iter 14749, loss: 0.5387 Iter 14750, loss: 0.4926 Iter 14751, loss: 0.4614 Iter 14752, loss: 0.5055 Iter 14753, loss: 0.5490 Iter 14754, loss: 0.4832 Iter 14755, loss: 0.5109 Iter 14756, loss: 0.5433 Iter 14757, loss: 0.4982 Iter 14758, loss: 0.5007 Iter 14759, loss: 0.5165 Iter 14760, loss: 0.5070 Iter 14761, loss: 0.5081 Iter 14762, loss: 0.4773 Iter 14763, loss: 0.4630 Iter 14764, loss: 0.4902 Iter 14765, loss: 0.5289 Iter 14766, loss: 0.4750 Iter 14767, loss: 0.4958 Iter 14768, loss: 0.4925 Iter 14769, loss: 0.4803 Iter 14770, loss: 0.5166 Iter 14771, loss: 0.5188 Iter 14772, loss: 0.4937 Iter 14773, loss: 0.4923 Iter 14774, loss: 0.4844 Iter 14775, loss: 0.4528 Iter 14776, loss: 0.4963 Iter 14777, loss: 0.4299 Iter 14778, loss: 0.4870 Iter 14779, loss: 0.5350 Iter 14780, loss: 0.5005 Iter 14781, loss: 0.4858 Iter 14782, loss: 0.4930 Iter 14783, loss: 0.4951 Iter 14784, loss: 0.5152 Iter 14785, loss: 0.4898 Iter 14786, loss: 0.5142

Iter 14787, loss: 0.4819 Iter 14788, loss: 0.4829 Iter 14789, loss: 0.4960 Iter 14790, loss: 0.5081 Iter 14791, loss: 0.5062 Iter 14792, loss: 0.5002 Iter 14793, loss: 0.4779 Iter 14794, loss: 0.5066 Iter 14795, loss: 0.4953 Iter 14796, loss: 0.4703 Iter 14797, loss: 0.4758 Iter 14798, loss: 0.4514 Iter 14799, loss: 0.4781 Iter 14800, loss: 0.4730 Iter 14801, loss: 0.4656 Iter 14802, loss: 0.4519 Iter 14803, loss: 0.4832 Iter 14804, loss: 0.5038 Iter 14805, loss: 0.4838 Iter 14806, loss: 0.4886 Iter 14807, loss: 0.4743 Iter 14808, loss: 0.5203 Iter 14809, loss: 0.4550 Iter 14810, loss: 0.5091 Iter 14811, loss: 0.4499 Iter 14812, loss: 0.4655 Iter 14813, loss: 0.4955 Iter 14814, loss: 0.4511 Iter 14815, loss: 0.4720 Iter 14816, loss: 0.4699 Iter 14817, loss: 0.5056 Iter 14818, loss: 0.4729 Iter 14819, loss: 0.4356 Iter 14820, loss: 0.4805 Iter 14821, loss: 0.4870 Iter 14822, loss: 0.4901 Iter 14823, loss: 0.4688 Iter 14824, loss: 0.4747 Iter 14825, loss: 0.4972 Iter 14826, loss: 0.4717 Iter 14827, loss: 0.4884 Iter 14828, loss: 0.4872

Iter 14829, loss: 0.4962 Iter 14830, loss: 0.4945 Iter 14831, loss: 0.4882 Iter 14832, loss: 0.4540 Iter 14833, loss: 0.4944 Iter 14834, loss: 0.5207 Iter 14835, loss: 0.4789 Iter 14836, loss: 0.4620 Iter 14837, loss: 0.4812 Iter 14838, loss: 0.4779 Iter 14839, loss: 0.4748 Iter 14840, loss: 0.4860 Iter 14841, loss: 0.4605 Iter 14842, loss: 0.4675 Iter 14843, loss: 0.4712 Iter 14844, loss: 0.5416 Iter 14845, loss: 0.5084 Iter 14846, loss: 0.4667 Iter 14847, loss: 0.4937 Iter 14848, loss: 0.4995 Iter 14849, loss: 0.4969 Iter 14850, loss: 0.5121 Iter 14851, loss: 0.4887 Iter 14852, loss: 0.4627 Iter 14853, loss: 0.4764 Iter 14854, loss: 0.4659 Iter 14855, loss: 0.4882 Iter 14856, loss: 0.4576 Iter 14857, loss: 0.4843 Iter 14858, loss: 0.4924 Iter 14859, loss: 0.4697 Iter 14860, loss: 0.4627 Iter 14861, loss: 0.4780 Iter 14862, loss: 0.4867 Iter 14863, loss: 0.4661 Iter 14864, loss: 0.4920 Iter 14865, loss: 0.4756 Iter 14866, loss: 0.4832 Iter 14867, loss: 0.4826 Iter 14868, loss: 0.4908 Iter 14869, loss: 0.4637 Iter 14870, loss: 0.5021

Iter 14871, loss: 0.4718 Iter 14872, loss: 0.5098 Iter 14873, loss: 0.5249 Iter 14874, loss: 0.4851 Iter 14875, loss: 0.4946 Iter 14876, loss: 0.4763 Iter 14877, loss: 0.4853 Iter 14878, loss: 0.4833 Iter 14879, loss: 0.5124 Iter 14880, loss: 0.4436 Iter 14881, loss: 0.4989 Iter 14882, loss: 0.4878 Iter 14883, loss: 0.4858 Iter 14884, loss: 0.4715 Iter 14885, loss: 0.4841 Iter 14886, loss: 0.4924 Iter 14887, loss: 0.5100 Iter 14888, loss: 0.4905 Iter 14889, loss: 0.4704 Iter 14890, loss: 0.5051 Iter 14891, loss: 0.5459 Iter 14892, loss: 0.5368 Iter 14893, loss: 0.5126 Iter 14894, loss: 0.4776 Iter 14895, loss: 0.4894 Iter 14896, loss: 0.4831 Iter 14897, loss: 0.4832 Iter 14898, loss: 0.4990 Iter 14899, loss: 0.4553 Iter 14900, loss: 0.5024 Iter 14901, loss: 0.5202 Iter 14902, loss: 0.5142 Iter 14903, loss: 0.5032 Iter 14904, loss: 0.5156 Iter 14905, loss: 0.4882 Iter 14906, loss: 0.5093 Iter 14907, loss: 0.4939 Iter 14908, loss: 0.5128 Iter 14909, loss: 0.4829 Iter 14910, loss: 0.4721 Iter 14911, loss: 0.4688 Iter 14912, loss: 0.4885

Iter 14913, loss: 0.4685 Iter 14914, loss: 0.4822 Iter 14915, loss: 0.4520 Iter 14916, loss: 0.4693 Iter 14917, loss: 0.4805 Iter 14918, loss: 0.4851 Iter 14919, loss: 0.5019 Iter 14920, loss: 0.4548 Iter 14921, loss: 0.4602 Iter 14922, loss: 0.4873 Iter 14923, loss: 0.5115 Iter 14924, loss: 0.5012 Iter 14925, loss: 0.4813 Iter 14926, loss: 0.5069 Iter 14927, loss: 0.5082 Iter 14928, loss: 0.4577 Iter 14929, loss: 0.5040 Iter 14930, loss: 0.4772 Iter 14931, loss: 0.5054 Iter 14932, loss: 0.5176 Iter 14933, loss: 0.4796 Iter 14934, loss: 0.4594 Iter 14935, loss: 0.4973 Iter 14936, loss: 0.4994 Iter 14937, loss: 0.5063 Iter 14938, loss: 0.5177 Iter 14939, loss: 0.4641 Iter 14940, loss: 0.5035 Iter 14941, loss: 0.4770 Iter 14942, loss: 0.4767 Iter 14943, loss: 0.4757 Iter 14944, loss: 0.4844 Iter 14945, loss: 0.4405 Iter 14946, loss: 0.4897 Iter 14947, loss: 0.4992 Iter 14948, loss: 0.4796 Iter 14949, loss: 0.4714 Iter 14950, loss: 0.4309 Iter 14951, loss: 0.4865 Iter 14952, loss: 0.4569 Iter 14953, loss: 0.4942

Iter 14954, loss: 0.5184 Iter 14955, loss: 0.4843 Iter 14956, loss: 0.5007 Iter 14957, loss: 0.4885 Iter 14958, loss: 0.5085 Iter 14959, loss: 0.4602 Iter 14960, loss: 0.4845 Iter 14961, loss: 0.5065 Iter 14962, loss: 0.4771 Iter 14963, loss: 0.4880 Iter 14964, loss: 0.4936 Iter 14965, loss: 0.5156 Iter 14966, loss: 0.4694 Iter 14967, loss: 0.4972 Iter 14968, loss: 0.4767 Iter 14969, loss: 0.4667 Iter 14970, loss: 0.4779 Iter 14971, loss: 0.4936 Iter 14972, loss: 0.4829 Iter 14973, loss: 0.4951 Iter 14974, loss: 0.5001 Iter 14975, loss: 0.5359 Iter 14976, loss: 0.4668 Iter 14977, loss: 0.5032 Iter 14978, loss: 0.4885 Iter 14979, loss: 0.5308 Iter 14980, loss: 0.4853 Iter 14981, loss: 0.4830 Iter 14982, loss: 0.4745 Iter 14983, loss: 0.4796 Iter 14984, loss: 0.4238 Iter 14985, loss: 0.4944 Iter 14986, loss: 0.4913 Iter 14987, loss: 0.4710 Iter 14988, loss: 0.4902 Iter 14989, loss: 0.5183 Iter 14990, loss: 0.5024 Iter 14991, loss: 0.4784 Iter 14992, loss: 0.4755 Iter 14993, loss: 0.4878 Iter 14994, loss: 0.4884 Iter 14995, loss: 0.5357

Iter 14996, loss: 0.4792 Iter 14997, loss: 0.4541 Iter 14998, loss: 0.4940 Iter 14999, loss: 0.4808 Iter 15000, loss: 0.4758 Iter 15001, loss: 0.4497 Iter 15002, loss: 0.4904 Iter 15003, loss: 0.4852 Iter 15004, loss: 0.4977 Iter 15005, loss: 0.4697 Iter 15006, loss: 0.4706 Iter 15007, loss: 0.5011 Iter 15008, loss: 0.5129 Iter 15009, loss: 0.5154 Iter 15010, loss: 0.4639 Iter 15011, loss: 0.5301 Iter 15012, loss: 0.4757 Iter 15013, loss: 0.4896 Iter 15014, loss: 0.4870 Iter 15015, loss: 0.5038 Iter 15016, loss: 0.4885 Iter 15017, loss: 0.4488 Iter 15018, loss: 0.4729 Iter 15019, loss: 0.4664 Iter 15020, loss: 0.4565 Iter 15021, loss: 0.4541 Iter 15022, loss: 0.4926 Iter 15023, loss: 0.5022 Iter 15024, loss: 0.5161 Iter 15025, loss: 0.4627 Iter 15026, loss: 0.4650 Iter 15027, loss: 0.5060 Iter 15028, loss: 0.5121 Iter 15029, loss: 0.4816 Iter 15030, loss: 0.4522 Iter 15031, loss: 0.4624 Iter 15032, loss: 0.4703 Iter 15033, loss: 0.4580 Iter 15034, loss: 0.4642 Iter 15035, loss: 0.4834 Iter 15036, loss: 0.4775 Iter 15037, loss: 0.5254

Iter 15038, loss: 0.4940 Iter 15039, loss: 0.4879 Iter 15040, loss: 0.4707 Iter 15041, loss: 0.4743 Iter 15042, loss: 0.4989 Iter 15043, loss: 0.4986 Iter 15044, loss: 0.4889 Iter 15045, loss: 0.4543 Iter 15046, loss: 0.4571 Iter 15047, loss: 0.4794 Iter 15048, loss: 0.4791 Iter 15049, loss: 0.4488 Iter 15050, loss: 0.5236 Iter 15051, loss: 0.5002 Iter 15052, loss: 0.4855 Iter 15053, loss: 0.4694 Iter 15054, loss: 0.5164 Iter 15055, loss: 0.4508 Iter 15056, loss: 0.4981 Iter 15057, loss: 0.5842 Iter 15058, loss: 0.5173 Iter 15059, loss: 0.5066 Iter 15060, loss: 0.4576 Iter 15061, loss: 0.4646 Iter 15062, loss: 0.5060 Iter 15063, loss: 0.5099 Iter 15064, loss: 0.4698 Iter 15065, loss: 0.4650 Iter 15066, loss: 0.5272 Iter 15067, loss: 0.4836 Iter 15068, loss: 0.5056 Iter 15069, loss: 0.4865 Iter 15070, loss: 0.4864 Iter 15071, loss: 0.4687 Iter 15072, loss: 0.5018 Iter 15073, loss: 0.5010 Iter 15074, loss: 0.5036 Iter 15075, loss: 0.4915 Iter 15076, loss: 0.4863 Iter 15077, loss: 0.4792 Iter 15078, loss: 0.5161 Iter 15079, loss: 0.4868

Iter 15080, loss: 0.5078 Iter 15081, loss: 0.5018 Iter 15082, loss: 0.5123 Iter 15083, loss: 0.4748 Iter 15084, loss: 0.4522 Iter 15085, loss: 0.4968 Iter 15086, loss: 0.4907 Iter 15087, loss: 0.4763 Iter 15088, loss: 0.4788 Iter 15089, loss: 0.5162 Iter 15090, loss: 0.4870 Iter 15091, loss: 0.5097 Iter 15092, loss: 0.4824 Iter 15093, loss: 0.4803 Iter 15094, loss: 0.5096 Iter 15095, loss: 0.5272 Iter 15096, loss: 0.4594 Iter 15097, loss: 0.4628 Iter 15098, loss: 0.4507 Iter 15099, loss: 0.4552 Iter 15100, loss: 0.5208 Iter 15101, loss: 0.4875 Iter 15102, loss: 0.4870 Iter 15103, loss: 0.4986 Iter 15104, loss: 0.5063 Iter 15105, loss: 0.5084 Iter 15106, loss: 0.4892 Iter 15107, loss: 0.4777 Iter 15108, loss: 0.4865 Iter 15109, loss: 0.4245 Iter 15110, loss: 0.4715 Iter 15111, loss: 0.4755 Iter 15112, loss: 0.4669 Iter 15113, loss: 0.4711 Iter 15114, loss: 0.5172 Iter 15115, loss: 0.4742 Iter 15116, loss: 0.4909 Iter 15117, loss: 0.4726 Iter 15118, loss: 0.4513 Iter 15119, loss: 0.4809 Iter 15120, loss: 0.4643 Iter 15121, loss: 0.4861

Iter 15122, loss: 0.4764 Iter 15123, loss: 0.5030 Iter 15124, loss: 0.5346 Iter 15125, loss: 0.4981 Iter 15126, loss: 0.5077 Iter 15127, loss: 0.4953 Iter 15128, loss: 0.5090 Iter 15129, loss: 0.4963 Iter 15130, loss: 0.4638 Iter 15131, loss: 0.4518 Iter 15132, loss: 0.4441 Iter 15133, loss: 0.4643 Iter 15134, loss: 0.4914 Iter 15135, loss: 0.4858 Iter 15136, loss: 0.4762 Iter 15137, loss: 0.4206 Iter 15138, loss: 0.4708 Iter 15139, loss: 0.4750 Iter 15140, loss: 0.5005 Iter 15141, loss: 0.4865 Iter 15142, loss: 0.4866 Iter 15143, loss: 0.4955 Iter 15144, loss: 0.4663 Iter 15145, loss: 0.5513 Iter 15146, loss: 0.4889 Iter 15147, loss: 0.4973 Iter 15148, loss: 0.4554 Iter 15149, loss: 0.5100 Iter 15150, loss: 0.4916 Iter 15151, loss: 0.5005 Iter 15152, loss: 0.4816 Iter 15153, loss: 0.4792 Iter 15154, loss: 0.4512 Iter 15155, loss: 0.4910 Iter 15156, loss: 0.4865 Iter 15157, loss: 0.4928 Iter 15158, loss: 0.4450 Iter 15159, loss: 0.4728 Iter 15160, loss: 0.4977 Iter 15161, loss: 0.4866 Iter 15162, loss: 0.4812 Iter 15163, loss: 0.4616

Iter 15164, loss: 0.4883 Iter 15165, loss: 0.4911 Iter 15166, loss: 0.4525 Iter 15167, loss: 0.4789 Iter 15168, loss: 0.5061 Iter 15169, loss: 0.5070 Iter 15170, loss: 0.4581 Iter 15171, loss: 0.5097 Iter 15172, loss: 0.5361 Iter 15173, loss: 0.4320 Iter 15174, loss: 0.5022 Iter 15175, loss: 0.4840 Iter 15176, loss: 0.4770 Iter 15177, loss: 0.4862 Iter 15178, loss: 0.5131 Iter 15179, loss: 0.4746 Iter 15180, loss: 0.4799 Iter 15181, loss: 0.4602 Iter 15182, loss: 0.5160 Iter 15183, loss: 0.4859 Iter 15184, loss: 0.4649 Iter 15185, loss: 0.4925 Iter 15186, loss: 0.5062 Iter 15187, loss: 0.5408 Iter 15188, loss: 0.4814 Iter 15189, loss: 0.5319 Iter 15190, loss: 0.4734 Iter 15191, loss: 0.4800 Iter 15192, loss: 0.4879 Iter 15193, loss: 0.5033 Iter 15194, loss: 0.4647 Iter 15195, loss: 0.5026 Iter 15196, loss: 0.5236 Iter 15197, loss: 0.4905 Iter 15198, loss: 0.4364 Iter 15199, loss: 0.4986 Iter 15200, loss: 0.4742 Iter 15201, loss: 0.4835 Iter 15202, loss: 0.4498 Iter 15203, loss: 0.4500 Iter 15204, loss: 0.4508 Iter 15205, loss: 0.5023

Iter 15206, loss: 0.4935 Iter 15207, loss: 0.4484 Iter 15208, loss: 0.4848 Iter 15209, loss: 0.5297 Iter 15210, loss: 0.4570 Iter 15211, loss: 0.4876 Iter 15212, loss: 0.4562 Iter 15213, loss: 0.4843 Iter 15214, loss: 0.5221 Iter 15215, loss: 0.5081 Iter 15216, loss: 0.4690 Iter 15217, loss: 0.4789 Iter 15218, loss: 0.4706 Iter 15219, loss: 0.4903 Iter 15220, loss: 0.4814 Iter 15221, loss: 0.5102 Iter 15222, loss: 0.5034 Iter 15223, loss: 0.4790 Iter 15224, loss: 0.4880 Iter 15225, loss: 0.4564 Iter 15226, loss: 0.4850 Iter 15227, loss: 0.4497 Iter 15228, loss: 0.4824 Iter 15229, loss: 0.4842 Iter 15230, loss: 0.4810 Iter 15231, loss: 0.5071 Iter 15232, loss: 0.4714 Iter 15233, loss: 0.4660 Iter 15234, loss: 0.5154 Iter 15235, loss: 0.4730 Iter 15236, loss: 0.5320 Iter 15237, loss: 0.4715 Iter 15238, loss: 0.4851 Iter 15239, loss: 0.4671 Iter 15240, loss: 0.4916 Iter 15241, loss: 0.5064 Iter 15242, loss: 0.4757 Iter 15243, loss: 0.4684 Iter 15244, loss: 0.4764 Iter 15245, loss: 0.4802 Iter 15246, loss: 0.4662 Iter 15247, loss: 0.4673

Iter 15248, loss: 0.4460 Iter 15249, loss: 0.5026 Iter 15250, loss: 0.4654 Iter 15251, loss: 0.5025 Iter 15252, loss: 0.5052 Iter 15253, loss: 0.4874 Iter 15254, loss: 0.4686 Iter 15255, loss: 0.4979 Iter 15256, loss: 0.4803 Iter 15257, loss: 0.4868 Iter 15258, loss: 0.4594 Iter 15259, loss: 0.4944 Iter 15260, loss: 0.4775 Iter 15261, loss: 0.4708 Iter 15262, loss: 0.4745 Iter 15263, loss: 0.4555 Iter 15264, loss: 0.4419 Iter 15265, loss: 0.4977 Iter 15266, loss: 0.5052 Iter 15267, loss: 0.4266 Iter 15268, loss: 0.4849 Iter 15269, loss: 0.4743 Iter 15270, loss: 0.4574 Iter 15271, loss: 0.4698 Iter 15272, loss: 0.4686 Iter 15273, loss: 0.5051 Iter 15274, loss: 0.5279 Iter 15275, loss: 0.4638 Iter 15276, loss: 0.4761 Iter 15277, loss: 0.5173 Iter 15278, loss: 0.4513 Iter 15279, loss: 0.4948 Iter 15280, loss: 0.5217 Iter 15281, loss: 0.4843 Iter 15282, loss: 0.4989 Iter 15283, loss: 0.4802 Iter 15284, loss: 0.5142 Iter 15285, loss: 0.4816 Iter 15286, loss: 0.5035 Iter 15287, loss: 0.4640 Iter 15288, loss: 0.4799 Iter 15289, loss: 0.4908

Iter 15290, loss: 0.4842 Iter 15291, loss: 0.4995 Iter 15292, loss: 0.4731 Iter 15293, loss: 0.4818 Iter 15294, loss: 0.4527 Iter 15295, loss: 0.4738 Iter 15296, loss: 0.4897 Iter 15297, loss: 0.4972 Iter 15298, loss: 0.5055 Iter 15299, loss: 0.5136 Iter 15300, loss: 0.4670 Iter 15301, loss: 0.4678 Iter 15302, loss: 0.4993 Iter 15303, loss: 0.5049 Iter 15304, loss: 0.4876 Iter 15305, loss: 0.4742 Iter 15306, loss: 0.5019 Iter 15307, loss: 0.4336 Iter 15308, loss: 0.4818 Iter 15309, loss: 0.4672 Iter 15310, loss: 0.5046 Iter 15311, loss: 0.5094 Iter 15312, loss: 0.4941 Iter 15313, loss: 0.4679 Iter 15314, loss: 0.4610 Iter 15315, loss: 0.5004 Iter 15316, loss: 0.5009 Iter 15317, loss: 0.4552 Iter 15318, loss: 0.5000 Iter 15319, loss: 0.4864 Iter 15320, loss: 0.4751 Iter 15321, loss: 0.5131 Iter 15322, loss: 0.4707 Iter 15323, loss: 0.4933 Iter 15324, loss: 0.5112 Iter 15325, loss: 0.4596 Iter 15326, loss: 0.4492 Iter 15327, loss: 0.5157 Iter 15328, loss: 0.4817 Iter 15329, loss: 0.4781 Iter 15330, loss: 0.4824 Iter 15331, loss: 0.4875

Iter 15332, loss: 0.4730 Iter 15333, loss: 0.4712 Iter 15334, loss: 0.4716 Iter 15335, loss: 0.4885 Iter 15336, loss: 0.4901 Iter 15337, loss: 0.4982 Iter 15338, loss: 0.4725 Iter 15339, loss: 0.5064 Iter 15340, loss: 0.4855 Iter 15341, loss: 0.4812 Iter 15342, loss: 0.4788 Iter 15343, loss: 0.5052 Iter 15344, loss: 0.4761 Iter 15345, loss: 0.5000 Iter 15346, loss: 0.4778 Iter 15347, loss: 0.4476 Iter 15348, loss: 0.5038 Iter 15349, loss: 0.4901 Iter 15350, loss: 0.4705 Iter 15351, loss: 0.4778 Iter 15352, loss: 0.4782 Iter 15353, loss: 0.4680 Iter 15354, loss: 0.4808 Iter 15355, loss: 0.4820 Iter 15356, loss: 0.4467 Iter 15357, loss: 0.4390 Iter 15358, loss: 0.4610 Iter 15359, loss: 0.4846 Iter 15360, loss: 0.4475 Iter 15361, loss: 0.4750 Iter 15362, loss: 0.4708 Iter 15363, loss: 0.5138 Iter 15364, loss: 0.4742 Iter 15365, loss: 0.4768 Iter 15366, loss: 0.4854 Iter 15367, loss: 0.5059 Iter 15368, loss: 0.4551 Iter 15369, loss: 0.4417 Iter 15370, loss: 0.4619 Iter 15371, loss: 0.4705 Iter 15372, loss: 0.4987 Iter 15373, loss: 0.5035

Iter 15374, loss: 0.5120 Iter 15375, loss: 0.4790 Iter 15376, loss: 0.4727 Iter 15377, loss: 0.5043 Iter 15378, loss: 0.4778 Iter 15379, loss: 0.5030 Iter 15380, loss: 0.4766 Iter 15381, loss: 0.5229 Iter 15382, loss: 0.4342 Iter 15383, loss: 0.4840 Iter 15384, loss: 0.4585 Iter 15385, loss: 0.4968 Iter 15386, loss: 0.4806 Iter 15387, loss: 0.5272 Iter 15388, loss: 0.4824 Iter 15389, loss: 0.4977 Iter 15390, loss: 0.5259 Iter 15391, loss: 0.5029 Iter 15392, loss: 0.5108 Iter 15393, loss: 0.4644 Iter 15394, loss: 0.4530 Iter 15395, loss: 0.5189 Iter 15396, loss: 0.4922 Iter 15397, loss: 0.4556 Iter 15398, loss: 0.5580 Iter 15399, loss: 0.4706 Iter 15400, loss: 0.4670 Iter 15401, loss: 0.4951 Iter 15402, loss: 0.4878 Iter 15403, loss: 0.4988 Iter 15404, loss: 0.4752 Iter 15405, loss: 0.4909 Iter 15406, loss: 0.4836 Iter 15407, loss: 0.4799 Iter 15408, loss: 0.4988 Iter 15409, loss: 0.4643 Iter 15410, loss: 0.5030 Iter 15411, loss: 0.4432 Iter 15412, loss: 0.4711 Iter 15413, loss: 0.4865 Iter 15414, loss: 0.4957 Iter 15415, loss: 0.4582

Iter 15416, loss: 0.5095 Iter 15417, loss: 0.5306 Iter 15418, loss: 0.5142 Iter 15419, loss: 0.4899 Iter 15420, loss: 0.4490 Iter 15421, loss: 0.4787 Iter 15422, loss: 0.4853 Iter 15423, loss: 0.4638 Iter 15424, loss: 0.4698 Iter 15425, loss: 0.4854 Iter 15426, loss: 0.4874 Iter 15427, loss: 0.5292 Iter 15428, loss: 0.4901 Iter 15429, loss: 0.4847 Iter 15430, loss: 0.4640 Iter 15431, loss: 0.4913 Iter 15432, loss: 0.5104 Iter 15433, loss: 0.4404 Iter 15434, loss: 0.4790 Iter 15435, loss: 0.5120 Iter 15436, loss: 0.5191 Iter 15437, loss: 0.5045 Iter 15438, loss: 0.4676 Iter 15439, loss: 0.4581 Iter 15440, loss: 0.4916 Iter 15441, loss: 0.4881 Iter 15442, loss: 0.4602 Iter 15443, loss: 0.4998 Iter 15444, loss: 0.4434 Iter 15445, loss: 0.5047 Iter 15446, loss: 0.4840 Iter 15447, loss: 0.4695 Iter 15448, loss: 0.5126 Iter 15449, loss: 0.4777 Iter 15450, loss: 0.4936 Iter 15451, loss: 0.4488 Iter 15452, loss: 0.4968 Iter 15453, loss: 0.4934 Iter 15454, loss: 0.4400 Iter 15455, loss: 0.4735 Iter 15456, loss: 0.4868 Iter 15457, loss: 0.5171

Iter 15458, loss: 0.4909 Iter 15459, loss: 0.4919 Iter 15460, loss: 0.4715 Iter 15461, loss: 0.4908 Iter 15462, loss: 0.4651 Iter 15463, loss: 0.5210 Iter 15464, loss: 0.5099 Iter 15465, loss: 0.4870 Iter 15466, loss: 0.4729 Iter 15467, loss: 0.4530 Iter 15468, loss: 0.4857 Iter 15469, loss: 0.4629 Iter 15470, loss: 0.5068 Iter 15471, loss: 0.5184 Iter 15472, loss: 0.4761 Iter 15473, loss: 0.4927 Iter 15474, loss: 0.4850 Iter 15475, loss: 0.4750 Iter 15476, loss: 0.4748 Iter 15477, loss: 0.5618 Iter 15478, loss: 0.4520 Iter 15479, loss: 0.4411 Iter 15480, loss: 0.4987 Iter 15481, loss: 0.4881 Iter 15482, loss: 0.4892 Iter 15483, loss: 0.4633 Iter 15484, loss: 0.4922 Iter 15485, loss: 0.4490 Iter 15486, loss: 0.4828 Iter 15487, loss: 0.4644 Iter 15488, loss: 0.4958 Iter 15489, loss: 0.4520 Iter 15490, loss: 0.4606 Iter 15491, loss: 0.4981 Iter 15492, loss: 0.4928 Iter 15493, loss: 0.4555 Iter 15494, loss: 0.4688 Iter 15495, loss: 0.5017 Iter 15496, loss: 0.4754 Iter 15497, loss: 0.4652 Iter 15498, loss: 0.4949 Iter 15499, loss: 0.4831

Iter 15500, loss: 0.4941 Iter 15501, loss: 0.4997 Iter 15502, loss: 0.4668 Iter 15503, loss: 0.4963 Iter 15504, loss: 0.5037 Iter 15505, loss: 0.4893 Iter 15506, loss: 0.4835 Iter 15507, loss: 0.5132 Iter 15508, loss: 0.4837 Iter 15509, loss: 0.5088 Iter 15510, loss: 0.4610 Iter 15511, loss: 0.5020 Iter 15512, loss: 0.5013 Iter 15513, loss: 0.4776 Iter 15514, loss: 0.4790 Iter 15515, loss: 0.4501 Iter 15516, loss: 0.4833 Iter 15517, loss: 0.4715 Iter 15518, loss: 0.4977 Iter 15519, loss: 0.4747 Iter 15520, loss: 0.4953 Iter 15521, loss: 0.4897 Iter 15522, loss: 0.4626 Iter 15523, loss: 0.5031 Iter 15524, loss: 0.5034 Iter 15525, loss: 0.4970 Iter 15526, loss: 0.4686 Iter 15527, loss: 0.5077 Iter 15528, loss: 0.4901 Iter 15529, loss: 0.4741 Iter 15530, loss: 0.4798 Iter 15531, loss: 0.4696 Iter 15532, loss: 0.5093 Iter 15533, loss: 0.4554 Iter 15534, loss: 0.4451 Iter 15535, loss: 0.4477 Iter 15536, loss: 0.5071 Iter 15537, loss: 0.5110 Iter 15538, loss: 0.5006 Iter 15539, loss: 0.4860 Iter 15540, loss: 0.5042 Iter 15541, loss: 0.4867

Iter 15542, loss: 0.4554 Iter 15543, loss: 0.4597 Iter 15544, loss: 0.4839 Iter 15545, loss: 0.4657 Iter 15546, loss: 0.4529 Iter 15547, loss: 0.4763 Iter 15548, loss: 0.5041 Iter 15549, loss: 0.4847 Iter 15550, loss: 0.5277 Iter 15551, loss: 0.5061 Iter 15552, loss: 0.4849 Iter 15553, loss: 0.4940 Iter 15554, loss: 0.4783 Iter 15555, loss: 0.4588 Iter 15556, loss: 0.4456 Iter 15557, loss: 0.4382 Iter 15558, loss: 0.4868 Iter 15559, loss: 0.5144 Iter 15560, loss: 0.4980 Iter 15561, loss: 0.5271 Iter 15562, loss: 0.4700 Iter 15563, loss: 0.4873 Iter 15564, loss: 0.5345 Iter 15565, loss: 0.5045 Iter 15566, loss: 0.4580 Iter 15567, loss: 0.4683 Iter 15568, loss: 0.4696 Iter 15569, loss: 0.5223 Iter 15570, loss: 0.4724 Iter 15571, loss: 0.4532 Iter 15572, loss: 0.4959 Iter 15573, loss: 0.4827 Iter 15574, loss: 0.5013 Iter 15575, loss: 0.4723 Iter 15576, loss: 0.4588 Iter 15577, loss: 0.5411 Iter 15578, loss: 0.4689 Iter 15579, loss: 0.5121 Iter 15580, loss: 0.4715 Iter 15581, loss: 0.4940 Iter 15582, loss: 0.4752 Iter 15583, loss: 0.5226

Iter 15584, loss: 0.4501 Iter 15585, loss: 0.4566 Iter 15586, loss: 0.4943 Iter 15587, loss: 0.4894 Iter 15588, loss: 0.5112 Iter 15589, loss: 0.4598 Iter 15590, loss: 0.4888 Iter 15591, loss: 0.4848 Iter 15592, loss: 0.4949 Iter 15593, loss: 0.4892 Iter 15594, loss: 0.5133 Iter 15595, loss: 0.5150 Iter 15596, loss: 0.5221 Iter 15597, loss: 0.4667 Iter 15598, loss: 0.4674 Iter 15599, loss: 0.5229 Iter 15600, loss: 0.4874 Iter 15601, loss: 0.4969 Iter 15602, loss: 0.4863 Iter 15603, loss: 0.5063 Iter 15604, loss: 0.5063 Iter 15605, loss: 0.4006 Iter 15606, loss: 0.4728 Iter 15607, loss: 0.5228 Iter 15608, loss: 0.5074 Iter 15609, loss: 0.4789 Iter 15610, loss: 0.4710 Iter 15611, loss: 0.5308 Iter 15612, loss: 0.4799 Iter 15613, loss: 0.4816 Iter 15614, loss: 0.4834 Iter 15615, loss: 0.4809 Iter 15616, loss: 0.5309 Iter 15617, loss: 0.4786 Iter 15618, loss: 0.5188 Iter 15619, loss: 0.4779 Iter 15620, loss: 0.4839 Iter 15621, loss: 0.4733 Iter 15622, loss: 0.5505 Iter 15623, loss: 0.4563 Iter 15624, loss: 0.4551 Iter 15625, loss: 0.4556

Iter 15626, loss: 0.4564 Iter 15627, loss: 0.4636 Iter 15628, loss: 0.4609 Iter 15629, loss: 0.4843 Iter 15630, loss: 0.4557 Iter 15631, loss: 0.4883 Iter 15632, loss: 0.4644 Iter 15633, loss: 0.5051 Iter 15634, loss: 0.4982 Iter 15635, loss: 0.4490 Iter 15636, loss: 0.4490 Iter 15637, loss: 0.4775 Iter 15638, loss: 0.4767 Iter 15639, loss: 0.4654 Iter 15640, loss: 0.4883 Iter 15641, loss: 0.5116 Iter 15642, loss: 0.5095 Iter 15643, loss: 0.4565 Iter 15644, loss: 0.4858 Iter 15645, loss: 0.4749 Iter 15646, loss: 0.4762 Iter 15647, loss: 0.4789 Iter 15648, loss: 0.5188 Iter 15649, loss: 0.4821 Iter 15650, loss: 0.5274 Iter 15651, loss: 0.4571 Iter 15652, loss: 0.4874 Iter 15653, loss: 0.4988 Iter 15654, loss: 0.4827 Iter 15655, loss: 0.4902 Iter 15656, loss: 0.4773 Iter 15657, loss: 0.4751 Iter 15658, loss: 0.4867 Iter 15659, loss: 0.5084 Iter 15660, loss: 0.4851 Iter 15661, loss: 0.4326 Iter 15662, loss: 0.4527 Iter 15663, loss: 0.4518 Iter 15664, loss: 0.4492 Iter 15665, loss: 0.5001 Iter 15666, loss: 0.4904 Iter 15667, loss: 0.4828

Iter 15668, loss: 0.4868 Iter 15669, loss: 0.4731 Iter 15670, loss: 0.4635 Iter 15671, loss: 0.4871 Iter 15672, loss: 0.4776 Iter 15673, loss: 0.4870 Iter 15674, loss: 0.4499 Iter 15675, loss: 0.4931 Iter 15676, loss: 0.4767 Iter 15677, loss: 0.4927 Iter 15678, loss: 0.4740 Iter 15679, loss: 0.4795 Iter 15680, loss: 0.5507 Iter 15681, loss: 0.4791 Iter 15682, loss: 0.4745 Iter 15683, loss: 0.5109 Iter 15684, loss: 0.4924 Iter 15685, loss: 0.4776 Iter 15686, loss: 0.4832 Iter 15687, loss: 0.4652 Iter 15688, loss: 0.4811 Iter 15689, loss: 0.4681 Iter 15690, loss: 0.4943 Iter 15691, loss: 0.4855 Iter 15692, loss: 0.5218 Iter 15693, loss: 0.5108 Iter 15694, loss: 0.4751 Iter 15695, loss: 0.4824 Iter 15696, loss: 0.4948 Iter 15697, loss: 0.4999 Iter 15698, loss: 0.4837 Iter 15699, loss: 0.4777 Iter 15700, loss: 0.4483 Iter 15701, loss: 0.4907 Iter 15702, loss: 0.4883 Iter 15703, loss: 0.4723 Iter 15704, loss: 0.4745 Iter 15705, loss: 0.4831 Iter 15706, loss: 0.4773 Iter 15707, loss: 0.5088 Iter 15708, loss: 0.4892 Iter 15709, loss: 0.4840

Iter 15710, loss: 0.4623 Iter 15711, loss: 0.4813 Iter 15712, loss: 0.4698 Iter 15713, loss: 0.4963 Iter 15714, loss: 0.5032 Iter 15715, loss: 0.4555 Iter 15716, loss: 0.4747 Iter 15717, loss: 0.5041 Iter 15718, loss: 0.4966 Iter 15719, loss: 0.4798 Iter 15720, loss: 0.4890 Iter 15721, loss: 0.4598 Iter 15722, loss: 0.4396 Iter 15723, loss: 0.4756 Iter 15724, loss: 0.5122 Iter 15725, loss: 0.5345 Iter 15726, loss: 0.5116 Iter 15727, loss: 0.5173 Iter 15728, loss: 0.4746 Iter 15729, loss: 0.5329 Iter 15730, loss: 0.4746 Iter 15731, loss: 0.5048 Iter 15732, loss: 0.4943 Iter 15733, loss: 0.4536 Iter 15734, loss: 0.4715 Iter 15735, loss: 0.4508 Iter 15736, loss: 0.4640 Iter 15737, loss: 0.4648 Iter 15738, loss: 0.4359 Iter 15739, loss: 0.5007 Iter 15740, loss: 0.5339 Iter 15741, loss: 0.4695 Iter 15742, loss: 0.4825 Iter 15743, loss: 0.4544 Iter 15744, loss: 0.4299 Iter 15745, loss: 0.4689 Iter 15746, loss: 0.4681 Iter 15747, loss: 0.5013 Iter 15748, loss: 0.4532 Iter 15749, loss: 0.4496 Iter 15750, loss: 0.4757

Iter 15751, loss: 0.4588 Iter 15752, loss: 0.5143 Iter 15753, loss: 0.4552 Iter 15754, loss: 0.4656 Iter 15755, loss: 0.4785 Iter 15756, loss: 0.5069 Iter 15757, loss: 0.4878 Iter 15758, loss: 0.4845 Iter 15759, loss: 0.4459 Iter 15760, loss: 0.4873 Iter 15761, loss: 0.5316 Iter 15762, loss: 0.4744 Iter 15763, loss: 0.5204 Iter 15764, loss: 0.4894 Iter 15765, loss: 0.4687 Iter 15766, loss: 0.5162 Iter 15767, loss: 0.5143 Iter 15768, loss: 0.5301 Iter 15769, loss: 0.4779 Iter 15770, loss: 0.4863 Iter 15771, loss: 0.4711 Iter 15772, loss: 0.4560 Iter 15773, loss: 0.4590 Iter 15774, loss: 0.5093 Iter 15775, loss: 0.4991 Iter 15776, loss: 0.4896 Iter 15777, loss: 0.4742 Iter 15778, loss: 0.4272 Iter 15779, loss: 0.4663 Iter 15780, loss: 0.4816 Iter 15781, loss: 0.4997 Iter 15782, loss: 0.4904 Iter 15783, loss: 0.5008 Iter 15784, loss: 0.4631 Iter 15785, loss: 0.4823 Iter 15786, loss: 0.4927 Iter 15787, loss: 0.4660 Iter 15788, loss: 0.4859 Iter 15789, loss: 0.4861 Iter 15790, loss: 0.5122 Iter 15791, loss: 0.4871 Iter 15792, loss: 0.4639

Iter 15793, loss: 0.5144 Iter 15794, loss: 0.4902 Iter 15795, loss: 0.4613 Iter 15796, loss: 0.4531 Iter 15797, loss: 0.5320 Iter 15798, loss: 0.4597 Iter 15799, loss: 0.5329 Iter 15800, loss: 0.4627 Iter 15801, loss: 0.4441 Iter 15802, loss: 0.5169 Iter 15803, loss: 0.4605 Iter 15804, loss: 0.4784 Iter 15805, loss: 0.4907 Iter 15806, loss: 0.4998 Iter 15807, loss: 0.5261 Iter 15808, loss: 0.4836 Iter 15809, loss: 0.4878 Iter 15810, loss: 0.5036 Iter 15811, loss: 0.4428 Iter 15812, loss: 0.4946 Iter 15813, loss: 0.4652 Iter 15814, loss: 0.4989 Iter 15815, loss: 0.4961 Iter 15816, loss: 0.4773 Iter 15817, loss: 0.4797 Iter 15818, loss: 0.4771 Iter 15819, loss: 0.5301 Iter 15820, loss: 0.4831 Iter 15821, loss: 0.4771 Iter 15822, loss: 0.5219 Iter 15823, loss: 0.4645 Iter 15824, loss: 0.4810 Iter 15825, loss: 0.5070 Iter 15826, loss: 0.4769 Iter 15827, loss: 0.4776 Iter 15828, loss: 0.4952 Iter 15829, loss: 0.4497 Iter 15830, loss: 0.5134 Iter 15831, loss: 0.4942 Iter 15832, loss: 0.4576 Iter 15833, loss: 0.5017 Iter 15834, loss: 0.4923

Iter 15835, loss: 0.4627 Iter 15836, loss: 0.4776 Iter 15837, loss: 0.4756 Iter 15838, loss: 0.4899 Iter 15839, loss: 0.4739 Iter 15840, loss: 0.4508 Iter 15841, loss: 0.4680 Iter 15842, loss: 0.4889 Iter 15843, loss: 0.4757 Iter 15844, loss: 0.4640 Iter 15845, loss: 0.4645 Iter 15846, loss: 0.4534 Iter 15847, loss: 0.5033 Iter 15848, loss: 0.4577 Iter 15849, loss: 0.4879 Iter 15850, loss: 0.4870 Iter 15851, loss: 0.4782 Iter 15852, loss: 0.4728 Iter 15853, loss: 0.4912 Iter 15854, loss: 0.4808 Iter 15855, loss: 0.4769 Iter 15856, loss: 0.4868 Iter 15857, loss: 0.4738 Iter 15858, loss: 0.4663 Iter 15859, loss: 0.4804 Iter 15860, loss: 0.5294 Iter 15861, loss: 0.4793 Iter 15862, loss: 0.4776 Iter 15863, loss: 0.4828 Iter 15864, loss: 0.4806 Iter 15865, loss: 0.4741 Iter 15866, loss: 0.5486 Iter 15867, loss: 0.4679 Iter 15868, loss: 0.4893 Iter 15869, loss: 0.4619 Iter 15870, loss: 0.4220 Iter 15871, loss: 0.4822 Iter 15872, loss: 0.4540 Iter 15873, loss: 0.4762 Iter 15874, loss: 0.4745 Iter 15875, loss: 0.4766 Iter 15876, loss: 0.4853

Iter 15877, loss: 0.5011 Iter 15878, loss: 0.5379 Iter 15879, loss: 0.5129 Iter 15880, loss: 0.4596 Iter 15881, loss: 0.4585 Iter 15882, loss: 0.4850 Iter 15883, loss: 0.4754 Iter 15884, loss: 0.4632 Iter 15885, loss: 0.4689 Iter 15886, loss: 0.5298 Iter 15887, loss: 0.5002 Iter 15888, loss: 0.4623 Iter 15889, loss: 0.4559 Iter 15890, loss: 0.4703 Iter 15891, loss: 0.4829 Iter 15892, loss: 0.5110 Iter 15893, loss: 0.5003 Iter 15894, loss: 0.4830 Iter 15895, loss: 0.5055 Iter 15896, loss: 0.4657 Iter 15897, loss: 0.4710 Iter 15898, loss: 0.4921 Iter 15899, loss: 0.4646 Iter 15900, loss: 0.4940 Iter 15901, loss: 0.4957 Iter 15902, loss: 0.4493 Iter 15903, loss: 0.5254 Iter 15904, loss: 0.5111 Iter 15905, loss: 0.4744 Iter 15906, loss: 0.4838 Iter 15907, loss: 0.4746 Iter 15908, loss: 0.4927 Iter 15909, loss: 0.4898 Iter 15910, loss: 0.5109 Iter 15911, loss: 0.4804 Iter 15912, loss: 0.4642 Iter 15913, loss: 0.4984 Iter 15914, loss: 0.4376 Iter 15915, loss: 0.4932 Iter 15916, loss: 0.4770 Iter 15917, loss: 0.4862 Iter 15918, loss: 0.4552

Iter 15919, loss: 0.5070 Iter 15920, loss: 0.5275 Iter 15921, loss: 0.4855 Iter 15922, loss: 0.4691 Iter 15923, loss: 0.5274 Iter 15924, loss: 0.5125 Iter 15925, loss: 0.4691 Iter 15926, loss: 0.5055 Iter 15927, loss: 0.4804 Iter 15928, loss: 0.4971 Iter 15929, loss: 0.5042 Iter 15930, loss: 0.4628 Iter 15931, loss: 0.5223 Iter 15932, loss: 0.4561 Iter 15933, loss: 0.4609 Iter 15934, loss: 0.4964 Iter 15935, loss: 0.4763 Iter 15936, loss: 0.4522 Iter 15937, loss: 0.4857 Iter 15938, loss: 0.4832 Iter 15939, loss: 0.4673 Iter 15940, loss: 0.5117 Iter 15941, loss: 0.4485 Iter 15942, loss: 0.4955 Iter 15943, loss: 0.5038 Iter 15944, loss: 0.5203 Iter 15945, loss: 0.4746 Iter 15946, loss: 0.4472 Iter 15947, loss: 0.5096 Iter 15948, loss: 0.4673 Iter 15949, loss: 0.4941 Iter 15950, loss: 0.4972 Iter 15951, loss: 0.5205 Iter 15952, loss: 0.5150 Iter 15953, loss: 0.4656 Iter 15954, loss: 0.4753 Iter 15955, loss: 0.4789 Iter 15956, loss: 0.5293 Iter 15957, loss: 0.4752 Iter 15958, loss: 0.5003 Iter 15959, loss: 0.4866 Iter 15960, loss: 0.4973

Iter 15961, loss: 0.4731 Iter 15962, loss: 0.4887 Iter 15963, loss: 0.4739 Iter 15964, loss: 0.4785 Iter 15965, loss: 0.4272 Iter 15966, loss: 0.4885 Iter 15967, loss: 0.4884 Iter 15968, loss: 0.4900 Iter 15969, loss: 0.4917 Iter 15970, loss: 0.4804 Iter 15971, loss: 0.4676 Iter 15972, loss: 0.5062 Iter 15973, loss: 0.4605 Iter 15974, loss: 0.4501 Iter 15975, loss: 0.5166 Iter 15976, loss: 0.4594 Iter 15977, loss: 0.4566 Iter 15978, loss: 0.4822 Iter 15979, loss: 0.4701 Iter 15980, loss: 0.4883 Iter 15981, loss: 0.4927 Iter 15982, loss: 0.4956 Iter 15983, loss: 0.4662 Iter 15984, loss: 0.4683 Iter 15985, loss: 0.4972 Iter 15986, loss: 0.5291 Iter 15987, loss: 0.4942 Iter 15988, loss: 0.4893 Iter 15989, loss: 0.4428 Iter 15990, loss: 0.5125 Iter 15991, loss: 0.4886 Iter 15992, loss: 0.5025 Iter 15993, loss: 0.5052 Iter 15994, loss: 0.5141 Iter 15995, loss: 0.4892 Iter 15996, loss: 0.4986 Iter 15997, loss: 0.4962 Iter 15998, loss: 0.4829 Iter 15999, loss: 0.5160 Iter 16000, loss: 0.4952 Iter 16001, loss: 0.4681 Iter 16002, loss: 0.4709

Iter 16003, loss: 0.4780 Iter 16004, loss: 0.5057 Iter 16005, loss: 0.4888 Iter 16006, loss: 0.4724 Iter 16007, loss: 0.4509 Iter 16008, loss: 0.4368 Iter 16009, loss: 0.4826 Iter 16010, loss: 0.5108 Iter 16011, loss: 0.4876 Iter 16012, loss: 0.4982 Iter 16013, loss: 0.4834 Iter 16014, loss: 0.5246 Iter 16015, loss: 0.5135 Iter 16016, loss: 0.4676 Iter 16017, loss: 0.4680 Iter 16018, loss: 0.4748 Iter 16019, loss: 0.4799 Iter 16020, loss: 0.5157 Iter 16021, loss: 0.4806 Iter 16022, loss: 0.5201 Iter 16023, loss: 0.4848 Iter 16024, loss: 0.4766 Iter 16025, loss: 0.4977 Iter 16026, loss: 0.5080 Iter 16027, loss: 0.5465 Iter 16028, loss: 0.4604 Iter 16029, loss: 0.5190 Iter 16030, loss: 0.4797 Iter 16031, loss: 0.4560 Iter 16032, loss: 0.4553 Iter 16033, loss: 0.4897 Iter 16034, loss: 0.4723 Iter 16035, loss: 0.5108 Iter 16036, loss: 0.4849 Iter 16037, loss: 0.5080 Iter 16038, loss: 0.4601 Iter 16039, loss: 0.4819 Iter 16040, loss: 0.5186 Iter 16041, loss: 0.4690 Iter 16042, loss: 0.5313 Iter 16043, loss: 0.4632 Iter 16044, loss: 0.5035

Iter 16045, loss: 0.4905 Iter 16046, loss: 0.4647 Iter 16047, loss: 0.5175 Iter 16048, loss: 0.5193 Iter 16049, loss: 0.5078 Iter 16050, loss: 0.4711 Iter 16051, loss: 0.5101 Iter 16052, loss: 0.4637 Iter 16053, loss: 0.4755 Iter 16054, loss: 0.4898 Iter 16055, loss: 0.4919 Iter 16056, loss: 0.4579 Iter 16057, loss: 0.4973 Iter 16058, loss: 0.4719 Iter 16059, loss: 0.4765 Iter 16060, loss: 0.4829 Iter 16061, loss: 0.4358 Iter 16062, loss: 0.4846 Iter 16063, loss: 0.4840 Iter 16064, loss: 0.4774 Iter 16065, loss: 0.5164 Iter 16066, loss: 0.4389 Iter 16067, loss: 0.4871 Iter 16068, loss: 0.4884 Iter 16069, loss: 0.4783 Iter 16070, loss: 0.4689 Iter 16071, loss: 0.4836 Iter 16072, loss: 0.4556 Iter 16073, loss: 0.4853 Iter 16074, loss: 0.4546 Iter 16075, loss: 0.4684 Iter 16076, loss: 0.4858 Iter 16077, loss: 0.4753 Iter 16078, loss: 0.4599 Iter 16079, loss: 0.4449 Iter 16080, loss: 0.4817 Iter 16081, loss: 0.4946 Iter 16082, loss: 0.4542 Iter 16083, loss: 0.5237 Iter 16084, loss: 0.4669 Iter 16085, loss: 0.5043 Iter 16086, loss: 0.4767

Iter 16087, loss: 0.4592 Iter 16088, loss: 0.4180 Iter 16089, loss: 0.4846 Iter 16090, loss: 0.5021 Iter 16091, loss: 0.4799 Iter 16092, loss: 0.4829 Iter 16093, loss: 0.4632 Iter 16094, loss: 0.4952 Iter 16095, loss: 0.4565 Iter 16096, loss: 0.5042 Iter 16097, loss: 0.5003 Iter 16098, loss: 0.4775 Iter 16099, loss: 0.5064 Iter 16100, loss: 0.4629 Iter 16101, loss: 0.4640 Iter 16102, loss: 0.5137 Iter 16103, loss: 0.4865 Iter 16104, loss: 0.5018 Iter 16105, loss: 0.4786 Iter 16106, loss: 0.4507 Iter 16107, loss: 0.4558 Iter 16108, loss: 0.4869 Iter 16109, loss: 0.4677 Iter 16110, loss: 0.4751 Iter 16111, loss: 0.4825 Iter 16112, loss: 0.4569 Iter 16113, loss: 0.4706 Iter 16114, loss: 0.4819 Iter 16115, loss: 0.4929 Iter 16116, loss: 0.4920 Iter 16117, loss: 0.4649 Iter 16118, loss: 0.4376 Iter 16119, loss: 0.4776 Iter 16120, loss: 0.5087 Iter 16121, loss: 0.4654 Iter 16122, loss: 0.4677 Iter 16123, loss: 0.4783 Iter 16124, loss: 0.5035 Iter 16125, loss: 0.4859 Iter 16126, loss: 0.4734 Iter 16127, loss: 0.4812 Iter 16128, loss: 0.5051

Iter 16129, loss: 0.5070 Iter 16130, loss: 0.4593 Iter 16131, loss: 0.4666 Iter 16132, loss: 0.4884 Iter 16133, loss: 0.4665 Iter 16134, loss: 0.4824 Iter 16135, loss: 0.4674 Iter 16136, loss: 0.4728 Iter 16137, loss: 0.4914 Iter 16138, loss: 0.4590 Iter 16139, loss: 0.4601 Iter 16140, loss: 0.4858 Iter 16141, loss: 0.4731 Iter 16142, loss: 0.4950 Iter 16143, loss: 0.4549 Iter 16144, loss: 0.4687 Iter 16145, loss: 0.4426 Iter 16146, loss: 0.5252 Iter 16147, loss: 0.4994 Iter 16148, loss: 0.4945 Iter 16149, loss: 0.4559 Iter 16150, loss: 0.4645 Iter 16151, loss: 0.4793 Iter 16152, loss: 0.4911 Iter 16153, loss: 0.4815 Iter 16154, loss: 0.4748 Iter 16155, loss: 0.4572 Iter 16156, loss: 0.5240 Iter 16157, loss: 0.5028 Iter 16158, loss: 0.4823 Iter 16159, loss: 0.4709 Iter 16160, loss: 0.5396 Iter 16161, loss: 0.4524 Iter 16162, loss: 0.4973 Iter 16163, loss: 0.4441 Iter 16164, loss: 0.4554 Iter 16165, loss: 0.4753 Iter 16166, loss: 0.4809 Iter 16167, loss: 0.4675 Iter 16168, loss: 0.5332 Iter 16169, loss: 0.4374 Iter 16170, loss: 0.4882

Iter 16171, loss: 0.4913 Iter 16172, loss: 0.4752 Iter 16173, loss: 0.4816 Iter 16174, loss: 0.4798 Iter 16175, loss: 0.4821 Iter 16176, loss: 0.4633 Iter 16177, loss: 0.4524 Iter 16178, loss: 0.4925 Iter 16179, loss: 0.4626 Iter 16180, loss: 0.4827 Iter 16181, loss: 0.5033 Iter 16182, loss: 0.4449 Iter 16183, loss: 0.4742 Iter 16184, loss: 0.4577 Iter 16185, loss: 0.4921 Iter 16186, loss: 0.4816 Iter 16187, loss: 0.4574 Iter 16188, loss: 0.4729 Iter 16189, loss: 0.5021 Iter 16190, loss: 0.4879 Iter 16191, loss: 0.4757 Iter 16192, loss: 0.4700 Iter 16193, loss: 0.4938 Iter 16194, loss: 0.4855 Iter 16195, loss: 0.4588 Iter 16196, loss: 0.4912 Iter 16197, loss: 0.4507 Iter 16198, loss: 0.4926 Iter 16199, loss: 0.4403 Iter 16200, loss: 0.5150 Iter 16201, loss: 0.4759 Iter 16202, loss: 0.5070 Iter 16203, loss: 0.4894 Iter 16204, loss: 0.4690 Iter 16205, loss: 0.4583 Iter 16206, loss: 0.4669 Iter 16207, loss: 0.4906 Iter 16208, loss: 0.4716 Iter 16209, loss: 0.5135 Iter 16210, loss: 0.5094 Iter 16211, loss: 0.4671 Iter 16212, loss: 0.4902

Iter 16213, loss: 0.4333 Iter 16214, loss: 0.4663 Iter 16215, loss: 0.4885 Iter 16216, loss: 0.4742 Iter 16217, loss: 0.4591 Iter 16218, loss: 0.4614 Iter 16219, loss: 0.4619 Iter 16220, loss: 0.4606 Iter 16221, loss: 0.5133 Iter 16222, loss: 0.4529 Iter 16223, loss: 0.5159 Iter 16224, loss: 0.5001 Iter 16225, loss: 0.4653 Iter 16226, loss: 0.5292 Iter 16227, loss: 0.4689 Iter 16228, loss: 0.4628 Iter 16229, loss: 0.4976 Iter 16230, loss: 0.5287 Iter 16231, loss: 0.4919 Iter 16232, loss: 0.4767 Iter 16233, loss: 0.4801 Iter 16234, loss: 0.5059 Iter 16235, loss: 0.4667 Iter 16236, loss: 0.4717 Iter 16237, loss: 0.4534 Iter 16238, loss: 0.5034 Iter 16239, loss: 0.4863 Iter 16240, loss: 0.4780 Iter 16241, loss: 0.4960 Iter 16242, loss: 0.4681 Iter 16243, loss: 0.5099 Iter 16244, loss: 0.4824 Iter 16245, loss: 0.4556 Iter 16246, loss: 0.4911 Iter 16247, loss: 0.4541 Iter 16248, loss: 0.5090 Iter 16249, loss: 0.4869 Iter 16250, loss: 0.4976 Iter 16251, loss: 0.4850 Iter 16252, loss: 0.5283 Iter 16253, loss: 0.4738 Iter 16254, loss: 0.4825

Iter 16255, loss: 0.4981 Iter 16256, loss: 0.5253 Iter 16257, loss: 0.5303 Iter 16258, loss: 0.5060 Iter 16259, loss: 0.4997 Iter 16260, loss: 0.4804 Iter 16261, loss: 0.4894 Iter 16262, loss: 0.4809 Iter 16263, loss: 0.4789 Iter 16264, loss: 0.4707 Iter 16265, loss: 0.5111 Iter 16266, loss: 0.4706 Iter 16267, loss: 0.4669 Iter 16268, loss: 0.4889 Iter 16269, loss: 0.4891 Iter 16270, loss: 0.4824 Iter 16271, loss: 0.4349 Iter 16272, loss: 0.4908 Iter 16273, loss: 0.4699 Iter 16274, loss: 0.5510 Iter 16275, loss: 0.5133 Iter 16276, loss: 0.4885 Iter 16277, loss: 0.4795 Iter 16278, loss: 0.4440 Iter 16279, loss: 0.5095 Iter 16280, loss: 0.4969 Iter 16281, loss: 0.4753 Iter 16282, loss: 0.4624 Iter 16283, loss: 0.4627 Iter 16284, loss: 0.4587 Iter 16285, loss: 0.4861 Iter 16286, loss: 0.4603 Iter 16287, loss: 0.4500 Iter 16288, loss: 0.4876 Iter 16289, loss: 0.4601 Iter 16290, loss: 0.4738 Iter 16291, loss: 0.5040 Iter 16292, loss: 0.4983 Iter 16293, loss: 0.4891 Iter 16294, loss: 0.5111 Iter 16295, loss: 0.4788 Iter 16296, loss: 0.4520

Iter 16297, loss: 0.4419 Iter 16298, loss: 0.5240 Iter 16299, loss: 0.5046 Iter 16300, loss: 0.4896 Iter 16301, loss: 0.4876 Iter 16302, loss: 0.4532 Iter 16303, loss: 0.4729 Iter 16304, loss: 0.4907 Iter 16305, loss: 0.4444 Iter 16306, loss: 0.5020 Iter 16307, loss: 0.4886 Iter 16308, loss: 0.4737 Iter 16309, loss: 0.4964 Iter 16310, loss: 0.4651 Iter 16311, loss: 0.4693 Iter 16312, loss: 0.4474 Iter 16313, loss: 0.5159 Iter 16314, loss: 0.5035 Iter 16315, loss: 0.4975 Iter 16316, loss: 0.4680 Iter 16317, loss: 0.5096 Iter 16318, loss: 0.5136 Iter 16319, loss: 0.4927 Iter 16320, loss: 0.5066 Iter 16321, loss: 0.5149 Iter 16322, loss: 0.4470 Iter 16323, loss: 0.5184 Iter 16324, loss: 0.4364 Iter 16325, loss: 0.4667 Iter 16326, loss: 0.4343 Iter 16327, loss: 0.4918 Iter 16328, loss: 0.5022 Iter 16329, loss: 0.5054 Iter 16330, loss: 0.4694 Iter 16331, loss: 0.4506 Iter 16332, loss: 0.4629 Iter 16333, loss: 0.4865 Iter 16334, loss: 0.5019 Iter 16335, loss: 0.4201 Iter 16336, loss: 0.4812 Iter 16337, loss: 0.4783 Iter 16338, loss: 0.5080

Iter 16339, loss: 0.4498 Iter 16340, loss: 0.4941 Iter 16341, loss: 0.4995 Iter 16342, loss: 0.4619 Iter 16343, loss: 0.5014 Iter 16344, loss: 0.4688 Iter 16345, loss: 0.4680 Iter 16346, loss: 0.4890 Iter 16347, loss: 0.4976 Iter 16348, loss: 0.4623 Iter 16349, loss: 0.5011 Iter 16350, loss: 0.4719 Iter 16351, loss: 0.4647 Iter 16352, loss: 0.4457 Iter 16353, loss: 0.4610 Iter 16354, loss: 0.4875 Iter 16355, loss: 0.5299 Iter 16356, loss: 0.4797 Iter 16357, loss: 0.4271 Iter 16358, loss: 0.4706 Iter 16359, loss: 0.4680 Iter 16360, loss: 0.4940 Iter 16361, loss: 0.4461 Iter 16362, loss: 0.5112 Iter 16363, loss: 0.4638 Iter 16364, loss: 0.4782 Iter 16365, loss: 0.4884 Iter 16366, loss: 0.4989 Iter 16367, loss: 0.5037 Iter 16368, loss: 0.4911 Iter 16369, loss: 0.4966 Iter 16370, loss: 0.4565 Iter 16371, loss: 0.4982 Iter 16372, loss: 0.4875 Iter 16373, loss: 0.4871 Iter 16374, loss: 0.5096 Iter 16375, loss: 0.4814 Iter 16376, loss: 0.4661 Iter 16377, loss: 0.4733 Iter 16378, loss: 0.4720 Iter 16379, loss: 0.4616 Iter 16380, loss: 0.4533

Iter 16381, loss: 0.4500 Iter 16382, loss: 0.4629 Iter 16383, loss: 0.4822 Iter 16384, loss: 0.4539 Iter 16385, loss: 0.4641 Iter 16386, loss: 0.4712 Iter 16387, loss: 0.4777 Iter 16388, loss: 0.4657 Iter 16389, loss: 0.5320 Iter 16390, loss: 0.4542 Iter 16391, loss: 0.4743 Iter 16392, loss: 0.4820 Iter 16393, loss: 0.4852 Iter 16394, loss: 0.5185 Iter 16395, loss: 0.4192 Iter 16396, loss: 0.4686 Iter 16397, loss: 0.5116 Iter 16398, loss: 0.5150 Iter 16399, loss: 0.4916 Iter 16400, loss: 0.4705 Iter 16401, loss: 0.4787 Iter 16402, loss: 0.4892 Iter 16403, loss: 0.4683 Iter 16404, loss: 0.5099 Iter 16405, loss: 0.4733 Iter 16406, loss: 0.5048 Iter 16407, loss: 0.4905 Iter 16408, loss: 0.5061 Iter 16409, loss: 0.4687 Iter 16410, loss: 0.4562 Iter 16411, loss: 0.4391 Iter 16412, loss: 0.4592 Iter 16413, loss: 0.4622 Iter 16414, loss: 0.4763 Iter 16415, loss: 0.4894 Iter 16416, loss: 0.4811 Iter 16417, loss: 0.4894 Iter 16418, loss: 0.4518 Iter 16419, loss: 0.4769 Iter 16420, loss: 0.4686 Iter 16421, loss: 0.4875 Iter 16422, loss: 0.4579

Iter 16423, loss: 0.5179 Iter 16424, loss: 0.4573 Iter 16425, loss: 0.4901 Iter 16426, loss: 0.4463 Iter 16427, loss: 0.5055 Iter 16428, loss: 0.4768 Iter 16429, loss: 0.4696 Iter 16430, loss: 0.4455 Iter 16431, loss: 0.4728 Iter 16432, loss: 0.4646 Iter 16433, loss: 0.4376 Iter 16434, loss: 0.4869 Iter 16435, loss: 0.4869 Iter 16436, loss: 0.4959 Iter 16437, loss: 0.4255 Iter 16438, loss: 0.4668 Iter 16439, loss: 0.4620 Iter 16440, loss: 0.4567 Iter 16441, loss: 0.4711 Iter 16442, loss: 0.4817 Iter 16443, loss: 0.4855 Iter 16444, loss: 0.5198 Iter 16445, loss: 0.4845 Iter 16446, loss: 0.4773 Iter 16447, loss: 0.4800 Iter 16448, loss: 0.4849 Iter 16449, loss: 0.4492 Iter 16450, loss: 0.4838 Iter 16451, loss: 0.5364 Iter 16452, loss: 0.4700 Iter 16453, loss: 0.4933 Iter 16454, loss: 0.4566 Iter 16455, loss: 0.4704 Iter 16456, loss: 0.4921 Iter 16457, loss: 0.4747 Iter 16458, loss: 0.4749 Iter 16459, loss: 0.4581 Iter 16460, loss: 0.4641 Iter 16461, loss: 0.4855 Iter 16462, loss: 0.4731 Iter 16463, loss: 0.4795 Iter 16464, loss: 0.4930

Iter 16465, loss: 0.4861 Iter 16466, loss: 0.4404 Iter 16467, loss: 0.4837 Iter 16468, loss: 0.5365 Iter 16469, loss: 0.5122 Iter 16470, loss: 0.4922 Iter 16471, loss: 0.4731 Iter 16472, loss: 0.4561 Iter 16473, loss: 0.4890 Iter 16474, loss: 0.4422 Iter 16475, loss: 0.4623 Iter 16476, loss: 0.4775 Iter 16477, loss: 0.4684 Iter 16478, loss: 0.4931 Iter 16479, loss: 0.4706 Iter 16480, loss: 0.4795 Iter 16481, loss: 0.4543 Iter 16482, loss: 0.4592 Iter 16483, loss: 0.5257 Iter 16484, loss: 0.5218 Iter 16485, loss: 0.4391 Iter 16486, loss: 0.4632 Iter 16487, loss: 0.4709 Iter 16488, loss: 0.5074 Iter 16489, loss: 0.4621 Iter 16490, loss: 0.5075 Iter 16491, loss: 0.4521 Iter 16492, loss: 0.4977 Iter 16493, loss: 0.4659 Iter 16494, loss: 0.4683 Iter 16495, loss: 0.4833 Iter 16496, loss: 0.4764 Iter 16497, loss: 0.5230 Iter 16498, loss: 0.4984 Iter 16499, loss: 0.4730 Iter 16500, loss: 0.5024 Iter 16501, loss: 0.4837 Iter 16502, loss: 0.4342 Iter 16503, loss: 0.4603 Iter 16504, loss: 0.4375 Iter 16505, loss: 0.4669 Iter 16506, loss: 0.4787

Iter 16507, loss: 0.4955 Iter 16508, loss: 0.4952 Iter 16509, loss: 0.5026 Iter 16510, loss: 0.4544 Iter 16511, loss: 0.4921 Iter 16512, loss: 0.4560 Iter 16513, loss: 0.5166 Iter 16514, loss: 0.4670 Iter 16515, loss: 0.4683 Iter 16516, loss: 0.4826 Iter 16517, loss: 0.4676 Iter 16518, loss: 0.4386 Iter 16519, loss: 0.4851 Iter 16520, loss: 0.5195 Iter 16521, loss: 0.4819 Iter 16522, loss: 0.5066 Iter 16523, loss: 0.4628 Iter 16524, loss: 0.5062 Iter 16525, loss: 0.5140 Iter 16526, loss: 0.4835 Iter 16527, loss: 0.5078 Iter 16528, loss: 0.4633 Iter 16529, loss: 0.4781 Iter 16530, loss: 0.5053 Iter 16531, loss: 0.4890 Iter 16532, loss: 0.4666 Iter 16533, loss: 0.4672 Iter 16534, loss: 0.4496 Iter 16535, loss: 0.4656 Iter 16536, loss: 0.4462 Iter 16537, loss: 0.5262 Iter 16538, loss: 0.4593 Iter 16539, loss: 0.4995 Iter 16540, loss: 0.4510 Iter 16541, loss: 0.5305 Iter 16542, loss: 0.4568 Iter 16543, loss: 0.4490 Iter 16544, loss: 0.4479 Iter 16545, loss: 0.4720 Iter 16546, loss: 0.4751 Iter 16547, loss: 0.4852 Iter 16548, loss: 0.4608

Iter 16549, loss: 0.4560 Iter 16550, loss: 0.4638 Iter 16551, loss: 0.5078 Iter 16552, loss: 0.4810 Iter 16553, loss: 0.5099 Iter 16554, loss: 0.4610 Iter 16555, loss: 0.4702 Iter 16556, loss: 0.4730 Iter 16557, loss: 0.4448 Iter 16558, loss: 0.4823 Iter 16559, loss: 0.4629 Iter 16560, loss: 0.4714 Iter 16561, loss: 0.5070 Iter 16562, loss: 0.4903 Iter 16563, loss: 0.4242 Iter 16564, loss: 0.4629 Iter 16565, loss: 0.4842 Iter 16566, loss: 0.4684 Iter 16567, loss: 0.4808 Iter 16568, loss: 0.4757 Iter 16569, loss: 0.4537 Iter 16570, loss: 0.5082 Iter 16571, loss: 0.4807 Iter 16572, loss: 0.5159 Iter 16573, loss: 0.4548 Iter 16574, loss: 0.4675 Iter 16575, loss: 0.4581 Iter 16576, loss: 0.4861 Iter 16577, loss: 0.4974 Iter 16578, loss: 0.4611 Iter 16579, loss: 0.4906 Iter 16580, loss: 0.4512 Iter 16581, loss: 0.4647 Iter 16582, loss: 0.5306 Iter 16583, loss: 0.4898 Iter 16584, loss: 0.4747 Iter 16585, loss: 0.4684 Iter 16586, loss: 0.4954 Iter 16587, loss: 0.4657 Iter 16588, loss: 0.5014 Iter 16589, loss: 0.4477 Iter 16590, loss: 0.4698

Iter 16591, loss: 0.4454 Iter 16592, loss: 0.4777 Iter 16593, loss: 0.4928 Iter 16594, loss: 0.4822 Iter 16595, loss: 0.4650 Iter 16596, loss: 0.4683 Iter 16597, loss: 0.5427 Iter 16598, loss: 0.4951 Iter 16599, loss: 0.5328 Iter 16600, loss: 0.4468 Iter 16601, loss: 0.4841 Iter 16602, loss: 0.4693 Iter 16603, loss: 0.4907 Iter 16604, loss: 0.4513 Iter 16605, loss: 0.5098 Iter 16606, loss: 0.4603 Iter 16607, loss: 0.4858 Iter 16608, loss: 0.4687 Iter 16609, loss: 0.4326 Iter 16610, loss: 0.5045 Iter 16611, loss: 0.4567 Iter 16612, loss: 0.5043 Iter 16613, loss: 0.4976 Iter 16614, loss: 0.4646 Iter 16615, loss: 0.4540 Iter 16616, loss: 0.5031 Iter 16617, loss: 0.4663 Iter 16618, loss: 0.4648 Iter 16619, loss: 0.5115 Iter 16620, loss: 0.4689 Iter 16621, loss: 0.4655 Iter 16622, loss: 0.4739 Iter 16623, loss: 0.4733 Iter 16624, loss: 0.4743 Iter 16625, loss: 0.4761 Iter 16626, loss: 0.4730 Iter 16627, loss: 0.4619 Iter 16628, loss: 0.4838 Iter 16629, loss: 0.4681 Iter 16630, loss: 0.4716 Iter 16631, loss: 0.5063 Iter 16632, loss: 0.4827

Iter 16633, loss: 0.4778 Iter 16634, loss: 0.4752 Iter 16635, loss: 0.4725 Iter 16636, loss: 0.4580 Iter 16637, loss: 0.4631 Iter 16638, loss: 0.4799 Iter 16639, loss: 0.5249 Iter 16640, loss: 0.4970 Iter 16641, loss: 0.4975 Iter 16642, loss: 0.4565 Iter 16643, loss: 0.4828 Iter 16644, loss: 0.4196 Iter 16645, loss: 0.4823 Iter 16646, loss: 0.5318 Iter 16647, loss: 0.4553 Iter 16648, loss: 0.5205 Iter 16649, loss: 0.4675 Iter 16650, loss: 0.5059 Iter 16651, loss: 0.4569 Iter 16652, loss: 0.5150 Iter 16653, loss: 0.4885 Iter 16654, loss: 0.4733 Iter 16655, loss: 0.4915 Iter 16656, loss: 0.4921 Iter 16657, loss: 0.4294 Iter 16658, loss: 0.5075 Iter 16659, loss: 0.4567 Iter 16660, loss: 0.4870 Iter 16661, loss: 0.5008 Iter 16662, loss: 0.4961 Iter 16663, loss: 0.4337 Iter 16664, loss: 0.4992 Iter 16665, loss: 0.4446 Iter 16666, loss: 0.5021 Iter 16667, loss: 0.4955 Iter 16668, loss: 0.4726 Iter 16669, loss: 0.5401 Iter 16670, loss: 0.5028 Iter 16671, loss: 0.4382 Iter 16672, loss: 0.5080 Iter 16673, loss: 0.4377 Iter 16674, loss: 0.4931

Iter 16675, loss: 0.4503 Iter 16676, loss: 0.4809 Iter 16677, loss: 0.4850 Iter 16678, loss: 0.4918 Iter 16679, loss: 0.5057 Iter 16680, loss: 0.4371 Iter 16681, loss: 0.4715 Iter 16682, loss: 0.5046 Iter 16683, loss: 0.5351 Iter 16684, loss: 0.5166 Iter 16685, loss: 0.4886 Iter 16686, loss: 0.4843 Iter 16687, loss: 0.4808 Iter 16688, loss: 0.4852 Iter 16689, loss: 0.5035 Iter 16690, loss: 0.5238 Iter 16691, loss: 0.4332 Iter 16692, loss: 0.4421 Iter 16693, loss: 0.5065 Iter 16694, loss: 0.4794 Iter 16695, loss: 0.4691 Iter 16696, loss: 0.4669 Iter 16697, loss: 0.4809 Iter 16698, loss: 0.4312 Iter 16699, loss: 0.4921 Iter 16700, loss: 0.4861 Iter 16701, loss: 0.4553 Iter 16702, loss: 0.4902 Iter 16703, loss: 0.4550 Iter 16704, loss: 0.4695 Iter 16705, loss: 0.4815 Iter 16706, loss: 0.4897 Iter 16707, loss: 0.4724 Iter 16708, loss: 0.4623 Iter 16709, loss: 0.5154 Iter 16710, loss: 0.4649 Iter 16711, loss: 0.5086 Iter 16712, loss: 0.4367 Iter 16713, loss: 0.4593 Iter 16714, loss: 0.5052 Iter 16715, loss: 0.4960 Iter 16716, loss: 0.4914

Iter 16717, loss: 0.4952 Iter 16718, loss: 0.4822 Iter 16719, loss: 0.4140 Iter 16720, loss: 0.4948 Iter 16721, loss: 0.4695 Iter 16722, loss: 0.4401 Iter 16723, loss: 0.4527 Iter 16724, loss: 0.4979 Iter 16725, loss: 0.4499 Iter 16726, loss: 0.4861 Iter 16727, loss: 0.4898 Iter 16728, loss: 0.4814 Iter 16729, loss: 0.4687 Iter 16730, loss: 0.4905 Iter 16731, loss: 0.4486 Iter 16732, loss: 0.4541 Iter 16733, loss: 0.4668 Iter 16734, loss: 0.5005 Iter 16735, loss: 0.5026 Iter 16736, loss: 0.4750 Iter 16737, loss: 0.4529 Iter 16738, loss: 0.4607 Iter 16739, loss: 0.4949 Iter 16740, loss: 0.4864 Iter 16741, loss: 0.4593 Iter 16742, loss: 0.5292 Iter 16743, loss: 0.4945 Iter 16744, loss: 0.4567 Iter 16745, loss: 0.4344 Iter 16746, loss: 0.4817 Iter 16747, loss: 0.5111 Iter 16748, loss: 0.4825 Iter 16749, loss: 0.4921 Iter 16750, loss: 0.4852 Iter 16751, loss: 0.4816 Iter 16752, loss: 0.4818 Iter 16753, loss: 0.4842 Iter 16754, loss: 0.4982 Iter 16755, loss: 0.4816 Iter 16756, loss: 0.5011 Iter 16757, loss: 0.5148 Iter 16758, loss: 0.4939

Iter 16759, loss: 0.4837 Iter 16760, loss: 0.4537 Iter 16761, loss: 0.4695 Iter 16762, loss: 0.4798 Iter 16763, loss: 0.4657 Iter 16764, loss: 0.4876 Iter 16765, loss: 0.4553 Iter 16766, loss: 0.4530 Iter 16767, loss: 0.5209 Iter 16768, loss: 0.4870 Iter 16769, loss: 0.4703 Iter 16770, loss: 0.4794 Iter 16771, loss: 0.5290 Iter 16772, loss: 0.5072 Iter 16773, loss: 0.4638 Iter 16774, loss: 0.4396 Iter 16775, loss: 0.5002 Iter 16776, loss: 0.4990 Iter 16777, loss: 0.4542 Iter 16778, loss: 0.4490 Iter 16779, loss: 0.4785 Iter 16780, loss: 0.4835 Iter 16781, loss: 0.4892 Iter 16782, loss: 0.4603 Iter 16783, loss: 0.4516 Iter 16784, loss: 0.4773 Iter 16785, loss: 0.4557 Iter 16786, loss: 0.4320 Iter 16787, loss: 0.4817 Iter 16788, loss: 0.4587 Iter 16789, loss: 0.5100 Iter 16790, loss: 0.4877 Iter 16791, loss: 0.4594 Iter 16792, loss: 0.4812 Iter 16793, loss: 0.4878 Iter 16794, loss: 0.4803 Iter 16795, loss: 0.4835 Iter 16796, loss: 0.4663 Iter 16797, loss: 0.5078 Iter 16798, loss: 0.4868 Iter 16799, loss: 0.4801 Iter 16800, loss: 0.5123

Iter 16801, loss: 0.4787 Iter 16802, loss: 0.4545 Iter 16803, loss: 0.4636 Iter 16804, loss: 0.4892 Iter 16805, loss: 0.4781 Iter 16806, loss: 0.4721 Iter 16807, loss: 0.4872 Iter 16808, loss: 0.4583 Iter 16809, loss: 0.5390 Iter 16810, loss: 0.5234 Iter 16811, loss: 0.5025 Iter 16812, loss: 0.4291 Iter 16813, loss: 0.4466 Iter 16814, loss: 0.4888 Iter 16815, loss: 0.4578 Iter 16816, loss: 0.4688 Iter 16817, loss: 0.4500 Iter 16818, loss: 0.4890 Iter 16819, loss: 0.4952 Iter 16820, loss: 0.4727 Iter 16821, loss: 0.5016 Iter 16822, loss: 0.4420 Iter 16823, loss: 0.4922 Iter 16824, loss: 0.4495 Iter 16825, loss: 0.4833 Iter 16826, loss: 0.5016 Iter 16827, loss: 0.4550 Iter 16828, loss: 0.4753 Iter 16829, loss: 0.5055 Iter 16830, loss: 0.4936 Iter 16831, loss: 0.5298 Iter 16832, loss: 0.5078 Iter 16833, loss: 0.4689 Iter 16834, loss: 0.5159 Iter 16835, loss: 0.4925 Iter 16836, loss: 0.4721 Iter 16837, loss: 0.4779 Iter 16838, loss: 0.4590 Iter 16839, loss: 0.4449 Iter 16840, loss: 0.5307 Iter 16841, loss: 0.4788 Iter 16842, loss: 0.4797

Iter 16843, loss: 0.4836 Iter 16844, loss: 0.4717 Iter 16845, loss: 0.4872 Iter 16846, loss: 0.4999 Iter 16847, loss: 0.4945 Iter 16848, loss: 0.5147 Iter 16849, loss: 0.4813 Iter 16850, loss: 0.4951 Iter 16851, loss: 0.4705 Iter 16852, loss: 0.4858 Iter 16853, loss: 0.4702 Iter 16854, loss: 0.4467 Iter 16855, loss: 0.4624 Iter 16856, loss: 0.4917 Iter 16857, loss: 0.4986 Iter 16858, loss: 0.5363 Iter 16859, loss: 0.4678 Iter 16860, loss: 0.4707 Iter 16861, loss: 0.4922 Iter 16862, loss: 0.4370 Iter 16863, loss: 0.5140 Iter 16864, loss: 0.4473 Iter 16865, loss: 0.4683 Iter 16866, loss: 0.4787 Iter 16867, loss: 0.4674 Iter 16868, loss: 0.4823 Iter 16869, loss: 0.4448 Iter 16870, loss: 0.4751 Iter 16871, loss: 0.5173 Iter 16872, loss: 0.4966 Iter 16873, loss: 0.5001 Iter 16874, loss: 0.4895 Iter 16875, loss: 0.4441 Iter 16876, loss: 0.4915 Iter 16877, loss: 0.4503 Iter 16878, loss: 0.4998 Iter 16879, loss: 0.4907 Iter 16880, loss: 0.5173 Iter 16881, loss: 0.4794 Iter 16882, loss: 0.4555 Iter 16883, loss: 0.5122 Iter 16884, loss: 0.4982

Iter 16885, loss: 0.4665 Iter 16886, loss: 0.5057 Iter 16887, loss: 0.4659 Iter 16888, loss: 0.4681 Iter 16889, loss: 0.4909 Iter 16890, loss: 0.5016 Iter 16891, loss: 0.5000 Iter 16892, loss: 0.4734 Iter 16893, loss: 0.4758 Iter 16894, loss: 0.4514 Iter 16895, loss: 0.4852 Iter 16896, loss: 0.4842 Iter 16897, loss: 0.4772 Iter 16898, loss: 0.5056 Iter 16899, loss: 0.4744 Iter 16900, loss: 0.4704 Iter 16901, loss: 0.4793 Iter 16902, loss: 0.4432 Iter 16903, loss: 0.4466 Iter 16904, loss: 0.4683 Iter 16905, loss: 0.4856 Iter 16906, loss: 0.4756 Iter 16907, loss: 0.4745 Iter 16908, loss: 0.4666 Iter 16909, loss: 0.4541 Iter 16910, loss: 0.5047 Iter 16911, loss: 0.4415 Iter 16912, loss: 0.5166 Iter 16913, loss: 0.4625 Iter 16914, loss: 0.4229 Iter 16915, loss: 0.5115 Iter 16916, loss: 0.4854 Iter 16917, loss: 0.4907 Iter 16918, loss: 0.4827 Iter 16919, loss: 0.4845 Iter 16920, loss: 0.5126 Iter 16921, loss: 0.4454 Iter 16922, loss: 0.4772 Iter 16923, loss: 0.4588 Iter 16924, loss: 0.4962 Iter 16925, loss: 0.4658 Iter 16926, loss: 0.4948

Iter 16927, loss: 0.5136 Iter 16928, loss: 0.4853 Iter 16929, loss: 0.4403 Iter 16930, loss: 0.4587 Iter 16931, loss: 0.4578 Iter 16932, loss: 0.4339 Iter 16933, loss: 0.4948 Iter 16934, loss: 0.5157 Iter 16935, loss: 0.4770 Iter 16936, loss: 0.4604 Iter 16937, loss: 0.4828 Iter 16938, loss: 0.4521 Iter 16939, loss: 0.4868 Iter 16940, loss: 0.5058 Iter 16941, loss: 0.5059 Iter 16942, loss: 0.4588 Iter 16943, loss: 0.4537 Iter 16944, loss: 0.4862 Iter 16945, loss: 0.4707 Iter 16946, loss: 0.4832 Iter 16947, loss: 0.4603 Iter 16948, loss: 0.4710 Iter 16949, loss: 0.4891 Iter 16950, loss: 0.4704 Iter 16951, loss: 0.4104 Iter 16952, loss: 0.4573 Iter 16953, loss: 0.4686 Iter 16954, loss: 0.4375 Iter 16955, loss: 0.4535 Iter 16956, loss: 0.5142 Iter 16957, loss: 0.5095 Iter 16958, loss: 0.4750 Iter 16959, loss: 0.5170 Iter 16960, loss: 0.4573 Iter 16961, loss: 0.4754 Iter 16962, loss: 0.4727 Iter 16963, loss: 0.4802 Iter 16964, loss: 0.4521 Iter 16965, loss: 0.4799 Iter 16966, loss: 0.4557 Iter 16967, loss: 0.4888 Iter 16968, loss: 0.4881

Iter 16969, loss: 0.4748 Iter 16970, loss: 0.5064 Iter 16971, loss: 0.4318 Iter 16972, loss: 0.4703 Iter 16973, loss: 0.5082 Iter 16974, loss: 0.4606 Iter 16975, loss: 0.4882 Iter 16976, loss: 0.4717 Iter 16977, loss: 0.4618 Iter 16978, loss: 0.4566 Iter 16979, loss: 0.5353 Iter 16980, loss: 0.4786 Iter 16981, loss: 0.4495 Iter 16982, loss: 0.4842 Iter 16983, loss: 0.4776 Iter 16984, loss: 0.4688 Iter 16985, loss: 0.4722 Iter 16986, loss: 0.4949 Iter 16987, loss: 0.4411 Iter 16988, loss: 0.4989 Iter 16989, loss: 0.4359 Iter 16990, loss: 0.4599 Iter 16991, loss: 0.5039 Iter 16992, loss: 0.4807 Iter 16993, loss: 0.4547 Iter 16994, loss: 0.4422 Iter 16995, loss: 0.4456 Iter 16996, loss: 0.4889 Iter 16997, loss: 0.5042 Iter 16998, loss: 0.4567 Iter 16999, loss: 0.4684 Iter 17000, loss: 0.4802 Iter 17001, loss: 0.5550 Iter 17002, loss: 0.4644 Iter 17003, loss: 0.4851 Iter 17004, loss: 0.4359 Iter 17005, loss: 0.4987 Iter 17006, loss: 0.4688 Iter 17007, loss: 0.4595 Iter 17008, loss: 0.4468 Iter 17009, loss: 0.4497 Iter 17010, loss: 0.4648

Iter 17011, loss: 0.4518 Iter 17012, loss: 0.4598 Iter 17013, loss: 0.5019 Iter 17014, loss: 0.5376 Iter 17015, loss: 0.4308 Iter 17016, loss: 0.4935 Iter 17017, loss: 0.4986 Iter 17018, loss: 0.4663 Iter 17019, loss: 0.4655 Iter 17020, loss: 0.4569 Iter 17021, loss: 0.4378 Iter 17022, loss: 0.4899 Iter 17023, loss: 0.5053 Iter 17024, loss: 0.4366 Iter 17025, loss: 0.4701 Iter 17026, loss: 0.4605 Iter 17027, loss: 0.5191 Iter 17028, loss: 0.4967 Iter 17029, loss: 0.4847 Iter 17030, loss: 0.4998 Iter 17031, loss: 0.4680 Iter 17032, loss: 0.4478 Iter 17033, loss: 0.4220 Iter 17034, loss: 0.4867 Iter 17035, loss: 0.4719 Iter 17036, loss: 0.4559 Iter 17037, loss: 0.4812 Iter 17038, loss: 0.4322 Iter 17039, loss: 0.4830 Iter 17040, loss: 0.4605 Iter 17041, loss: 0.5373 Iter 17042, loss: 0.4904 Iter 17043, loss: 0.4686 Iter 17044, loss: 0.4791 Iter 17045, loss: 0.5288 Iter 17046, loss: 0.4610 Iter 17047, loss: 0.4929 Iter 17048, loss: 0.4684 Iter 17049, loss: 0.4689 Iter 17050, loss: 0.4402 Iter 17051, loss: 0.4400 Iter 17052, loss: 0.4923

Iter 17053, loss: 0.4542 Iter 17054, loss: 0.4408 Iter 17055, loss: 0.4684 Iter 17056, loss: 0.5011 Iter 17057, loss: 0.4751 Iter 17058, loss: 0.4799 Iter 17059, loss: 0.4735 Iter 17060, loss: 0.4786 Iter 17061, loss: 0.4905 Iter 17062, loss: 0.4653 Iter 17063, loss: 0.5110 Iter 17064, loss: 0.4540 Iter 17065, loss: 0.5038 Iter 17066, loss: 0.4567 Iter 17067, loss: 0.4771 Iter 17068, loss: 0.4769 Iter 17069, loss: 0.4734 Iter 17070, loss: 0.4975 Iter 17071, loss: 0.4788 Iter 17072, loss: 0.4815 Iter 17073, loss: 0.4768 Iter 17074, loss: 0.4906 Iter 17075, loss: 0.4728 Iter 17076, loss: 0.4687 Iter 17077, loss: 0.4537 Iter 17078, loss: 0.5356 Iter 17079, loss: 0.4927 Iter 17080, loss: 0.4680 Iter 17081, loss: 0.5039 Iter 17082, loss: 0.4376 Iter 17083, loss: 0.4603 Iter 17084, loss: 0.4792 Iter 17085, loss: 0.4862 Iter 17086, loss: 0.5030 Iter 17087, loss: 0.4984 Iter 17088, loss: 0.4676 Iter 17089, loss: 0.4698 Iter 17090, loss: 0.4725 Iter 17091, loss: 0.4565 Iter 17092, loss: 0.4836 Iter 17093, loss: 0.4624 Iter 17094, loss: 0.4562

Iter 17095, loss: 0.4636 Iter 17096, loss: 0.4963 Iter 17097, loss: 0.4788 Iter 17098, loss: 0.5004 Iter 17099, loss: 0.4692 Iter 17100, loss: 0.5033 Iter 17101, loss: 0.4850 Iter 17102, loss: 0.4615 Iter 17103, loss: 0.4981 Iter 17104, loss: 0.4862 Iter 17105, loss: 0.4396 Iter 17106, loss: 0.4660 Iter 17107, loss: 0.4781 Iter 17108, loss: 0.4656 Iter 17109, loss: 0.4580 Iter 17110, loss: 0.4918 Iter 17111, loss: 0.4671 Iter 17112, loss: 0.4750 Iter 17113, loss: 0.4482 Iter 17114, loss: 0.4870 Iter 17115, loss: 0.5286 Iter 17116, loss: 0.4497 Iter 17117, loss: 0.4518 Iter 17118, loss: 0.5310 Iter 17119, loss: 0.5040 Iter 17120, loss: 0.4789 Iter 17121, loss: 0.4664 Iter 17122, loss: 0.4715 Iter 17123, loss: 0.5034 Iter 17124, loss: 0.4522 Iter 17125, loss: 0.4760 Iter 17126, loss: 0.4608 Iter 17127, loss: 0.4899 Iter 17128, loss: 0.4803 Iter 17129, loss: 0.4385 Iter 17130, loss: 0.4585 Iter 17131, loss: 0.4744 Iter 17132, loss: 0.4885 Iter 17133, loss: 0.4971 Iter 17134, loss: 0.4790 Iter 17135, loss: 0.4860 Iter 17136, loss: 0.4727

Iter 17137, loss: 0.4519 Iter 17138, loss: 0.4705 Iter 17139, loss: 0.4533 Iter 17140, loss: 0.5085 Iter 17141, loss: 0.4515 Iter 17142, loss: 0.4786 Iter 17143, loss: 0.4653 Iter 17144, loss: 0.5110 Iter 17145, loss: 0.4816 Iter 17146, loss: 0.4593 Iter 17147, loss: 0.5075 Iter 17148, loss: 0.4292 Iter 17149, loss: 0.5038 Iter 17150, loss: 0.4870 Iter 17151, loss: 0.4750 Iter 17152, loss: 0.4639 Iter 17153, loss: 0.4724 Iter 17154, loss: 0.4865 Iter 17155, loss: 0.4921 Iter 17156, loss: 0.4323 Iter 17157, loss: 0.5003 Iter 17158, loss: 0.4629 Iter 17159, loss: 0.4754 Iter 17160, loss: 0.4647 Iter 17161, loss: 0.5430 Iter 17162, loss: 0.4726 Iter 17163, loss: 0.4592 Iter 17164, loss: 0.4717 Iter 17165, loss: 0.4618 Iter 17166, loss: 0.5033 Iter 17167, loss: 0.4512 Iter 17168, loss: 0.4759 Iter 17169, loss: 0.4643 Iter 17170, loss: 0.4924 Iter 17171, loss: 0.4888 Iter 17172, loss: 0.4929 Iter 17173, loss: 0.4715 Iter 17174, loss: 0.4653 Iter 17175, loss: 0.4793 Iter 17176, loss: 0.4836 Iter 17177, loss: 0.4751 Iter 17178, loss: 0.4966

Iter 17179, loss: 0.4824 Iter 17180, loss: 0.5047 Iter 17181, loss: 0.4577 Iter 17182, loss: 0.4728 Iter 17183, loss: 0.4809 Iter 17184, loss: 0.4351 Iter 17185, loss: 0.4774 Iter 17186, loss: 0.4946 Iter 17187, loss: 0.4688 Iter 17188, loss: 0.4753 Iter 17189, loss: 0.4887 Iter 17190, loss: 0.4430 Iter 17191, loss: 0.4784 Iter 17192, loss: 0.4918 Iter 17193, loss: 0.4519 Iter 17194, loss: 0.4736 Iter 17195, loss: 0.5100 Iter 17196, loss: 0.4349 Iter 17197, loss: 0.5003 Iter 17198, loss: 0.4765 Iter 17199, loss: 0.5064 Iter 17200, loss: 0.4467 Iter 17201, loss: 0.4942 Iter 17202, loss: 0.4435 Iter 17203, loss: 0.4503 Iter 17204, loss: 0.4703 Iter 17205, loss: 0.5147 Iter 17206, loss: 0.4699 Iter 17207, loss: 0.4875 Iter 17208, loss: 0.4723 Iter 17209, loss: 0.5008 Iter 17210, loss: 0.4504 Iter 17211, loss: 0.4750 Iter 17212, loss: 0.4809 Iter 17213, loss: 0.4451 Iter 17214, loss: 0.4708 Iter 17215, loss: 0.4737 Iter 17216, loss: 0.4464 Iter 17217, loss: 0.4827 Iter 17218, loss: 0.4451 Iter 17219, loss: 0.4753 Iter 17220, loss: 0.4941

Iter 17221, loss: 0.4886 Iter 17222, loss: 0.4823 Iter 17223, loss: 0.4860 Iter 17224, loss: 0.4724 Iter 17225, loss: 0.4495 Iter 17226, loss: 0.4916 Iter 17227, loss: 0.4851 Iter 17228, loss: 0.4830 Iter 17229, loss: 0.4867 Iter 17230, loss: 0.4462 Iter 17231, loss: 0.4356 Iter 17232, loss: 0.4666 Iter 17233, loss: 0.4944 Iter 17234, loss: 0.4813 Iter 17235, loss: 0.4865 Iter 17236, loss: 0.4811 Iter 17237, loss: 0.4622 Iter 17238, loss: 0.5092 Iter 17239, loss: 0.4702 Iter 17240, loss: 0.4910 Iter 17241, loss: 0.4769 Iter 17242, loss: 0.4722 Iter 17243, loss: 0.4558 Iter 17244, loss: 0.5005 Iter 17245, loss: 0.4745 Iter 17246, loss: 0.4832 Iter 17247, loss: 0.4810 Iter 17248, loss: 0.4987 Iter 17249, loss: 0.4870 Iter 17250, loss: 0.4823 Iter 17251, loss: 0.4630 Iter 17252, loss: 0.4593 Iter 17253, loss: 0.4858 Iter 17254, loss: 0.4972 Iter 17255, loss: 0.4685 Iter 17256, loss: 0.4895 Iter 17257, loss: 0.4889 Iter 17258, loss: 0.4842 Iter 17259, loss: 0.5057 Iter 17260, loss: 0.4914 Iter 17261, loss: 0.4999 Iter 17262, loss: 0.4934

Iter 17263, loss: 0.4617 Iter 17264, loss: 0.4527 Iter 17265, loss: 0.4664 Iter 17266, loss: 0.4685 Iter 17267, loss: 0.4705 Iter 17268, loss: 0.5040 Iter 17269, loss: 0.4751 Iter 17270, loss: 0.5032 Iter 17271, loss: 0.4580 Iter 17272, loss: 0.4415 Iter 17273, loss: 0.4981 Iter 17274, loss: 0.4992 Iter 17275, loss: 0.4895 Iter 17276, loss: 0.4599 Iter 17277, loss: 0.4183 Iter 17278, loss: 0.5152 Iter 17279, loss: 0.4570 Iter 17280, loss: 0.4708 Iter 17281, loss: 0.4600 Iter 17282, loss: 0.4937 Iter 17283, loss: 0.4826 Iter 17284, loss: 0.4679 Iter 17285, loss: 0.4884 Iter 17286, loss: 0.4849 Iter 17287, loss: 0.4810 Iter 17288, loss: 0.4965 Iter 17289, loss: 0.4762 Iter 17290, loss: 0.4638 Iter 17291, loss: 0.4398 Iter 17292, loss: 0.5023 Iter 17293, loss: 0.5026 Iter 17294, loss: 0.4979 Iter 17295, loss: 0.5010 Iter 17296, loss: 0.4267 Iter 17297, loss: 0.4475 Iter 17298, loss: 0.4763 Iter 17299, loss: 0.4941 Iter 17300, loss: 0.4886 Iter 17301, loss: 0.4655 Iter 17302, loss: 0.4288 Iter 17303, loss: 0.4429 Iter 17304, loss: 0.4538

Iter 17305, loss: 0.4928 Iter 17306, loss: 0.4573 Iter 17307, loss: 0.4512 Iter 17308, loss: 0.5069 Iter 17309, loss: 0.4358 Iter 17310, loss: 0.4433 Iter 17311, loss: 0.4656 Iter 17312, loss: 0.4339 Iter 17313, loss: 0.5007 Iter 17314, loss: 0.4739 Iter 17315, loss: 0.4758 Iter 17316, loss: 0.4868 Iter 17317, loss: 0.4944 Iter 17318, loss: 0.4645 Iter 17319, loss: 0.5079 Iter 17320, loss: 0.4998 Iter 17321, loss: 0.4861 Iter 17322, loss: 0.4640 Iter 17323, loss: 0.4508 Iter 17324, loss: 0.4758 Iter 17325, loss: 0.4845 Iter 17326, loss: 0.4658 Iter 17327, loss: 0.4639 Iter 17328, loss: 0.4350 Iter 17329, loss: 0.5026 Iter 17330, loss: 0.4805 Iter 17331, loss: 0.5089 Iter 17332, loss: 0.4521 Iter 17333, loss: 0.4542 Iter 17334, loss: 0.4544 Iter 17335, loss: 0.5048 Iter 17336, loss: 0.5182 Iter 17337, loss: 0.4869 Iter 17338, loss: 0.4832 Iter 17339, loss: 0.4697 Iter 17340, loss: 0.4666 Iter 17341, loss: 0.4988 Iter 17342, loss: 0.4757 Iter 17343, loss: 0.4270 Iter 17344, loss: 0.5151 Iter 17345, loss: 0.4593 Iter 17346, loss: 0.4577

Iter 17347, loss: 0.5024 Iter 17348, loss: 0.5302 Iter 17349, loss: 0.4806 Iter 17350, loss: 0.4947 Iter 17351, loss: 0.4652 Iter 17352, loss: 0.4767 Iter 17353, loss: 0.4536 Iter 17354, loss: 0.4780 Iter 17355, loss: 0.4325 Iter 17356, loss: 0.4614 Iter 17357, loss: 0.4626 Iter 17358, loss: 0.4393 Iter 17359, loss: 0.5102 Iter 17360, loss: 0.4692 Iter 17361, loss: 0.4633 Iter 17362, loss: 0.4916 Iter 17363, loss: 0.4712 Iter 17364, loss: 0.4788 Iter 17365, loss: 0.4488 Iter 17366, loss: 0.4762 Iter 17367, loss: 0.4548 Iter 17368, loss: 0.4742 Iter 17369, loss: 0.4682 Iter 17370, loss: 0.4718 Iter 17371, loss: 0.4946 Iter 17372, loss: 0.4843 Iter 17373, loss: 0.4882 Iter 17374, loss: 0.4731 Iter 17375, loss: 0.5038 Iter 17376, loss: 0.4761 Iter 17377, loss: 0.4665 Iter 17378, loss: 0.4723 Iter 17379, loss: 0.4751 Iter 17380, loss: 0.5107 Iter 17381, loss: 0.4289 Iter 17382, loss: 0.4856 Iter 17383, loss: 0.4806 Iter 17384, loss: 0.5060 Iter 17385, loss: 0.4766 Iter 17386, loss: 0.4365 Iter 17387, loss: 0.4752 Iter 17388, loss: 0.4928

Iter 17389, loss: 0.5011 Iter 17390, loss: 0.4814 Iter 17391, loss: 0.4702 Iter 17392, loss: 0.4720 Iter 17393, loss: 0.4759 Iter 17394, loss: 0.4402 Iter 17395, loss: 0.4534 Iter 17396, loss: 0.4504 Iter 17397, loss: 0.4592 Iter 17398, loss: 0.4572 Iter 17399, loss: 0.4896 Iter 17400, loss: 0.4872 Iter 17401, loss: 0.4671 Iter 17402, loss: 0.4795 Iter 17403, loss: 0.4918 Iter 17404, loss: 0.5113 Iter 17405, loss: 0.4420 Iter 17406, loss: 0.5092 Iter 17407, loss: 0.4772 Iter 17408, loss: 0.4813 Iter 17409, loss: 0.5040 Iter 17410, loss: 0.4850 Iter 17411, loss: 0.4859 Iter 17412, loss: 0.5199 Iter 17413, loss: 0.4485 Iter 17414, loss: 0.4648 Iter 17415, loss: 0.5082 Iter 17416, loss: 0.4873 Iter 17417, loss: 0.4896 Iter 17418, loss: 0.4366 Iter 17419, loss: 0.4431 Iter 17420, loss: 0.4902 Iter 17421, loss: 0.4900 Iter 17422, loss: 0.5035 Iter 17423, loss: 0.5063 Iter 17424, loss: 0.4324 Iter 17425, loss: 0.5039 Iter 17426, loss: 0.4639 Iter 17427, loss: 0.4979 Iter 17428, loss: 0.4517 Iter 17429, loss: 0.5084 Iter 17430, loss: 0.4319

Iter 17431, loss: 0.4503 Iter 17432, loss: 0.4461 Iter 17433, loss: 0.4573 Iter 17434, loss: 0.4981 Iter 17435, loss: 0.5020 Iter 17436, loss: 0.4477 Iter 17437, loss: 0.4331 Iter 17438, loss: 0.4487 Iter 17439, loss: 0.4943 Iter 17440, loss: 0.4309 Iter 17441, loss: 0.4585 Iter 17442, loss: 0.4654 Iter 17443, loss: 0.4298 Iter 17444, loss: 0.4551 Iter 17445, loss: 0.4728 Iter 17446, loss: 0.4322 Iter 17447, loss: 0.4817 Iter 17448, loss: 0.4901 Iter 17449, loss: 0.5317 Iter 17450, loss: 0.4793 Iter 17451, loss: 0.4125 Iter 17452, loss: 0.4780 Iter 17453, loss: 0.5212 Iter 17454, loss: 0.5185 Iter 17455, loss: 0.4443 Iter 17456, loss: 0.4879 Iter 17457, loss: 0.4802 Iter 17458, loss: 0.4263 Iter 17459, loss: 0.4728 Iter 17460, loss: 0.4548 Iter 17461, loss: 0.4633 Iter 17462, loss: 0.4813 Iter 17463, loss: 0.4862 Iter 17464, loss: 0.4652 Iter 17465, loss: 0.5253 Iter 17466, loss: 0.4549 Iter 17467, loss: 0.4425 Iter 17468, loss: 0.4587 Iter 17469, loss: 0.4815 Iter 17470, loss: 0.5362 Iter 17471, loss: 0.4504 Iter 17472, loss: 0.4835

Iter 17473, loss: 0.4720 Iter 17474, loss: 0.4762 Iter 17475, loss: 0.4528 Iter 17476, loss: 0.4706 Iter 17477, loss: 0.5398 Iter 17478, loss: 0.4545 Iter 17479, loss: 0.4802 Iter 17480, loss: 0.4882 Iter 17481, loss: 0.4976 Iter 17482, loss: 0.4405 Iter 17483, loss: 0.4597 Iter 17484, loss: 0.4579 Iter 17485, loss: 0.4623 Iter 17486, loss: 0.4539 Iter 17487, loss: 0.5030 Iter 17488, loss: 0.5006 Iter 17489, loss: 0.4757 Iter 17490, loss: 0.4733 Iter 17491, loss: 0.5151 Iter 17492, loss: 0.4836 Iter 17493, loss: 0.4551 Iter 17494, loss: 0.4522 Iter 17495, loss: 0.4768 Iter 17496, loss: 0.4957 Iter 17497, loss: 0.5080 Iter 17498, loss: 0.4696 Iter 17499, loss: 0.4988 Iter 17500, loss: 0.4496 Iter 17501, loss: 0.4437 Iter 17502, loss: 0.4618 Iter 17503, loss: 0.5014 Iter 17504, loss: 0.4941 Iter 17505, loss: 0.4941 Iter 17506, loss: 0.4862 Iter 17507, loss: 0.4579 Iter 17508, loss: 0.5133 Iter 17509, loss: 0.4782 Iter 17510, loss: 0.4619 Iter 17511, loss: 0.4552 Iter 17512, loss: 0.4591 Iter 17513, loss: 0.4635 Iter 17514, loss: 0.5139

Iter 17515, loss: 0.4765 Iter 17516, loss: 0.4860 Iter 17517, loss: 0.4713 Iter 17518, loss: 0.4572 Iter 17519, loss: 0.4661 Iter 17520, loss: 0.4513 Iter 17521, loss: 0.4576 Iter 17522, loss: 0.4328 Iter 17523, loss: 0.4360 Iter 17524, loss: 0.4430 Iter 17525, loss: 0.4564 Iter 17526, loss: 0.5237 Iter 17527, loss: 0.4731 Iter 17528, loss: 0.4817 Iter 17529, loss: 0.4484 Iter 17530, loss: 0.5051 Iter 17531, loss: 0.4944 Iter 17532, loss: 0.4771 Iter 17533, loss: 0.5382 Iter 17534, loss: 0.4504 Iter 17535, loss: 0.4800 Iter 17536, loss: 0.4827 Iter 17537, loss: 0.4558 Iter 17538, loss: 0.4623 Iter 17539, loss: 0.4818 Iter 17540, loss: 0.4756 Iter 17541, loss: 0.4402 Iter 17542, loss: 0.4459 Iter 17543, loss: 0.4695 Iter 17544, loss: 0.4982 Iter 17545, loss: 0.4659 Iter 17546, loss: 0.5408 Iter 17547, loss: 0.4690 Iter 17548, loss: 0.5526 Iter 17549, loss: 0.5093 Iter 17550, loss: 0.5027 Iter 17551, loss: 0.4996 Iter 17552, loss: 0.4557 Iter 17553, loss: 0.4813 Iter 17554, loss: 0.4627 Iter 17555, loss: 0.4849 Iter 17556, loss: 0.4901

Iter 17557, loss: 0.4874 Iter 17558, loss: 0.4519 Iter 17559, loss: 0.4785 Iter 17560, loss: 0.4800 Iter 17561, loss: 0.5131 Iter 17562, loss: 0.4707 Iter 17563, loss: 0.4754 Iter 17564, loss: 0.4311 Iter 17565, loss: 0.4919 Iter 17566, loss: 0.4417 Iter 17567, loss: 0.4746 Iter 17568, loss: 0.4833 Iter 17569, loss: 0.4742 Iter 17570, loss: 0.4457 Iter 17571, loss: 0.5193 Iter 17572, loss: 0.4997 Iter 17573, loss: 0.4798 Iter 17574, loss: 0.5015 Iter 17575, loss: 0.4856 Iter 17576, loss: 0.4410 Iter 17577, loss: 0.4870 Iter 17578, loss: 0.4750 Iter 17579, loss: 0.4879 Iter 17580, loss: 0.4790 Iter 17581, loss: 0.4959 Iter 17582, loss: 0.4646 Iter 17583, loss: 0.4986 Iter 17584, loss: 0.4871 Iter 17585, loss: 0.4810 Iter 17586, loss: 0.4728 Iter 17587, loss: 0.4729 Iter 17588, loss: 0.4913 Iter 17589, loss: 0.4698 Iter 17590, loss: 0.5003 Iter 17591, loss: 0.4914 Iter 17592, loss: 0.4674 Iter 17593, loss: 0.4666 Iter 17594, loss: 0.4609 Iter 17595, loss: 0.4547 Iter 17596, loss: 0.5012 Iter 17597, loss: 0.4560 Iter 17598, loss: 0.4796

Iter 17599, loss: 0.4500 Iter 17600, loss: 0.4450 Iter 17601, loss: 0.4626 Iter 17602, loss: 0.4670 Iter 17603, loss: 0.4657 Iter 17604, loss: 0.4681 Iter 17605, loss: 0.5073 Iter 17606, loss: 0.4488 Iter 17607, loss: 0.4860 Iter 17608, loss: 0.4542 Iter 17609, loss: 0.5495 Iter 17610, loss: 0.4701 Iter 17611, loss: 0.4895 Iter 17612, loss: 0.4930 Iter 17613, loss: 0.4846 Iter 17614, loss: 0.4880 Iter 17615, loss: 0.5356 Iter 17616, loss: 0.5112 Iter 17617, loss: 0.4664 Iter 17618, loss: 0.4999 Iter 17619, loss: 0.4711 Iter 17620, loss: 0.4834 Iter 17621, loss: 0.4981 Iter 17622, loss: 0.5018 Iter 17623, loss: 0.4476 Iter 17624, loss: 0.4995 Iter 17625, loss: 0.4526 Iter 17626, loss: 0.4776 Iter 17627, loss: 0.4771 Iter 17628, loss: 0.4612 Iter 17629, loss: 0.4991 Iter 17630, loss: 0.4862 Iter 17631, loss: 0.4967 Iter 17632, loss: 0.4976 Iter 17633, loss: 0.4354 Iter 17634, loss: 0.4880 Iter 17635, loss: 0.4965 Iter 17636, loss: 0.5205 Iter 17637, loss: 0.4893 Iter 17638, loss: 0.4581 Iter 17639, loss: 0.4562 Iter 17640, loss: 0.4989

Iter 17641, loss: 0.4924 Iter 17642, loss: 0.4862 Iter 17643, loss: 0.4760 Iter 17644, loss: 0.4615 Iter 17645, loss: 0.4854 Iter 17646, loss: 0.4404 Iter 17647, loss: 0.4892 Iter 17648, loss: 0.4626 Iter 17649, loss: 0.4756 Iter 17650, loss: 0.5097 Iter 17651, loss: 0.4769 Iter 17652, loss: 0.5107 Iter 17653, loss: 0.4666 Iter 17654, loss: 0.4632 Iter 17655, loss: 0.4332 Iter 17656, loss: 0.5027 Iter 17657, loss: 0.4501 Iter 17658, loss: 0.4918 Iter 17659, loss: 0.4828 Iter 17660, loss: 0.4687 Iter 17661, loss: 0.4726 Iter 17662, loss: 0.4714 Iter 17663, loss: 0.4711 Iter 17664, loss: 0.4986 Iter 17665, loss: 0.4772 Iter 17666, loss: 0.5305 Iter 17667, loss: 0.4717 Iter 17668, loss: 0.5246 Iter 17669, loss: 0.4911 Iter 17670, loss: 0.4651 Iter 17671, loss: 0.4486 Iter 17672, loss: 0.4257 Iter 17673, loss: 0.4799 Iter 17674, loss: 0.4906 Iter 17675, loss: 0.4730 Iter 17676, loss: 0.4400 Iter 17677, loss: 0.4442 Iter 17678, loss: 0.4970 Iter 17679, loss: 0.4559 Iter 17680, loss: 0.4688 Iter 17681, loss: 0.4188 Iter 17682, loss: 0.4393

Iter 17683, loss: 0.4976 Iter 17684, loss: 0.4879 Iter 17685, loss: 0.4915 Iter 17686, loss: 0.5230 Iter 17687, loss: 0.5009 Iter 17688, loss: 0.4652 Iter 17689, loss: 0.4107 Iter 17690, loss: 0.4585 Iter 17691, loss: 0.4753 Iter 17692, loss: 0.4716 Iter 17693, loss: 0.4610 Iter 17694, loss: 0.4798 Iter 17695, loss: 0.5035 Iter 17696, loss: 0.4649 Iter 17697, loss: 0.5049 Iter 17698, loss: 0.4570 Iter 17699, loss: 0.4431 Iter 17700, loss: 0.4598 Iter 17701, loss: 0.4501 Iter 17702, loss: 0.4674 Iter 17703, loss: 0.5043 Iter 17704, loss: 0.4734 Iter 17705, loss: 0.4651 Iter 17706, loss: 0.4965 Iter 17707, loss: 0.4578 Iter 17708, loss: 0.4675 Iter 17709, loss: 0.4924 Iter 17710, loss: 0.4855 Iter 17711, loss: 0.4800 Iter 17712, loss: 0.5182 Iter 17713, loss: 0.4956 Iter 17714, loss: 0.4792 Iter 17715, loss: 0.4938 Iter 17716, loss: 0.4432 Iter 17717, loss: 0.4982 Iter 17718, loss: 0.4779 Iter 17719, loss: 0.5172 Iter 17720, loss: 0.4602 Iter 17721, loss: 0.4757 Iter 17722, loss: 0.4643 Iter 17723, loss: 0.4633 Iter 17724, loss: 0.4497

Iter 17725, loss: 0.4302 Iter 17726, loss: 0.4891 Iter 17727, loss: 0.4365 Iter 17728, loss: 0.5136 Iter 17729, loss: 0.4817 Iter 17730, loss: 0.4817 Iter 17731, loss: 0.4794 Iter 17732, loss: 0.5104 Iter 17733, loss: 0.5131 Iter 17734, loss: 0.4712 Iter 17735, loss: 0.5185 Iter 17736, loss: 0.4997 Iter 17737, loss: 0.4987 Iter 17738, loss: 0.5117 Iter 17739, loss: 0.4776 Iter 17740, loss: 0.4724 Iter 17741, loss: 0.4487 Iter 17742, loss: 0.4863 Iter 17743, loss: 0.5137 Iter 17744, loss: 0.4982 Iter 17745, loss: 0.4572 Iter 17746, loss: 0.4703 Iter 17747, loss: 0.4906 Iter 17748, loss: 0.4901 Iter 17749, loss: 0.4756 Iter 17750, loss: 0.4818 Iter 17751, loss: 0.4965 Iter 17752, loss: 0.4509 Iter 17753, loss: 0.4237 Iter 17754, loss: 0.4830 Iter 17755, loss: 0.4852 Iter 17756, loss: 0.4744 Iter 17757, loss: 0.4750 Iter 17758, loss: 0.5036 Iter 17759, loss: 0.4687 Iter 17760, loss: 0.4498 Iter 17761, loss: 0.5142 Iter 17762, loss: 0.5156 Iter 17763, loss: 0.4635 Iter 17764, loss: 0.4869 Iter 17765, loss: 0.4568 Iter 17766, loss: 0.4425

Iter 17767, loss: 0.4556 Iter 17768, loss: 0.4743 Iter 17769, loss: 0.4692 Iter 17770, loss: 0.4978 Iter 17771, loss: 0.5346 Iter 17772, loss: 0.4568 Iter 17773, loss: 0.4602 Iter 17774, loss: 0.5025 Iter 17775, loss: 0.4480 Iter 17776, loss: 0.4236 Iter 17777, loss: 0.4626 Iter 17778, loss: 0.4666 Iter 17779, loss: 0.4766 Iter 17780, loss: 0.4563 Iter 17781, loss: 0.4529 Iter 17782, loss: 0.4871 Iter 17783, loss: 0.4836 Iter 17784, loss: 0.4885 Iter 17785, loss: 0.4603 Iter 17786, loss: 0.4672 Iter 17787, loss: 0.4517 Iter 17788, loss: 0.4307 Iter 17789, loss: 0.5204 Iter 17790, loss: 0.5040 Iter 17791, loss: 0.4531 Iter 17792, loss: 0.5144 Iter 17793, loss: 0.4536 Iter 17794, loss: 0.4773 Iter 17795, loss: 0.4580 Iter 17796, loss: 0.4993 Iter 17797, loss: 0.4468 Iter 17798, loss: 0.4622 Iter 17799, loss: 0.4808 Iter 17800, loss: 0.4725 Iter 17801, loss: 0.4748 Iter 17802, loss: 0.4788 Iter 17803, loss: 0.4887 Iter 17804, loss: 0.4720 Iter 17805, loss: 0.4737 Iter 17806, loss: 0.5041 Iter 17807, loss: 0.4572 Iter 17808, loss: 0.4684

Iter 17809, loss: 0.4405 Iter 17810, loss: 0.4479 Iter 17811, loss: 0.4829 Iter 17812, loss: 0.4533 Iter 17813, loss: 0.4743 Iter 17814, loss: 0.4735 Iter 17815, loss: 0.4689 Iter 17816, loss: 0.4721 Iter 17817, loss: 0.4752 Iter 17818, loss: 0.4760 Iter 17819, loss: 0.4588 Iter 17820, loss: 0.4752 Iter 17821, loss: 0.4933 Iter 17822, loss: 0.4837 Iter 17823, loss: 0.4739 Iter 17824, loss: 0.4836 Iter 17825, loss: 0.4599 Iter 17826, loss: 0.5013 Iter 17827, loss: 0.4367 Iter 17828, loss: 0.4899 Iter 17829, loss: 0.4512 Iter 17830, loss: 0.4826 Iter 17831, loss: 0.4680 Iter 17832, loss: 0.4629 Iter 17833, loss: 0.4822 Iter 17834, loss: 0.4637 Iter 17835, loss: 0.4556 Iter 17836, loss: 0.4805 Iter 17837, loss: 0.4897 Iter 17838, loss: 0.4923 Iter 17839, loss: 0.5057 Iter 17840, loss: 0.4676 Iter 17841, loss: 0.4474 Iter 17842, loss: 0.4465 Iter 17843, loss: 0.4587 Iter 17844, loss: 0.5415 Iter 17845, loss: 0.4796 Iter 17846, loss: 0.4759 Iter 17847, loss: 0.4939 Iter 17848, loss: 0.4731 Iter 17849, loss: 0.4362 Iter 17850, loss: 0.5129

Iter 17851, loss: 0.5072 Iter 17852, loss: 0.5075 Iter 17853, loss: 0.5413 Iter 17854, loss: 0.4707 Iter 17855, loss: 0.4679 Iter 17856, loss: 0.4551 Iter 17857, loss: 0.4829 Iter 17858, loss: 0.5033 Iter 17859, loss: 0.4969 Iter 17860, loss: 0.4836 Iter 17861, loss: 0.4817 Iter 17862, loss: 0.5064 Iter 17863, loss: 0.4992 Iter 17864, loss: 0.4724 Iter 17865, loss: 0.4850 Iter 17866, loss: 0.4645 Iter 17867, loss: 0.4621 Iter 17868, loss: 0.4315 Iter 17869, loss: 0.4727 Iter 17870, loss: 0.4551 Iter 17871, loss: 0.4752 Iter 17872, loss: 0.4913 Iter 17873, loss: 0.4576 Iter 17874, loss: 0.4536 Iter 17875, loss: 0.4766 Iter 17876, loss: 0.5169 Iter 17877, loss: 0.4644 Iter 17878, loss: 0.4810 Iter 17879, loss: 0.5111 Iter 17880, loss: 0.4892 Iter 17881, loss: 0.4665 Iter 17882, loss: 0.4498 Iter 17883, loss: 0.4378 Iter 17884, loss: 0.4565 Iter 17885, loss: 0.5014 Iter 17886, loss: 0.5015 Iter 17887, loss: 0.4816 Iter 17888, loss: 0.4712 Iter 17889, loss: 0.4681 Iter 17890, loss: 0.4891 Iter 17891, loss: 0.4965 Iter 17892, loss: 0.5027

Iter 17893, loss: 0.4669 Iter 17894, loss: 0.5088 Iter 17895, loss: 0.4687 Iter 17896, loss: 0.4352 Iter 17897, loss: 0.5205 Iter 17898, loss: 0.4704 Iter 17899, loss: 0.4577 Iter 17900, loss: 0.5126 Iter 17901, loss: 0.5232 Iter 17902, loss: 0.4470 Iter 17903, loss: 0.4729 Iter 17904, loss: 0.4896 Iter 17905, loss: 0.4983 Iter 17906, loss: 0.5036 Iter 17907, loss: 0.4987 Iter 17908, loss: 0.4674 Iter 17909, loss: 0.4427 Iter 17910, loss: 0.4853 Iter 17911, loss: 0.5265 Iter 17912, loss: 0.4826 Iter 17913, loss: 0.4590 Iter 17914, loss: 0.5127 Iter 17915, loss: 0.5049 Iter 17916, loss: 0.4783 Iter 17917, loss: 0.4537 Iter 17918, loss: 0.5214 Iter 17919, loss: 0.4682 Iter 17920, loss: 0.5055 Iter 17921, loss: 0.4683 Iter 17922, loss: 0.4685 Iter 17923, loss: 0.5096 Iter 17924, loss: 0.4707 Iter 17925, loss: 0.4739 Iter 17926, loss: 0.4825 Iter 17927, loss: 0.4458 Iter 17928, loss: 0.4846 Iter 17929, loss: 0.4738 Iter 17930, loss: 0.5053 Iter 17931, loss: 0.4748 Iter 17932, loss: 0.4559 Iter 17933, loss: 0.4670

Iter 17934, loss: 0.4480 Iter 17935, loss: 0.4808 Iter 17936, loss: 0.4686 Iter 17937, loss: 0.4958 Iter 17938, loss: 0.4887 Iter 17939, loss: 0.4990 Iter 17940, loss: 0.4838 Iter 17941, loss: 0.5096 Iter 17942, loss: 0.5058 Iter 17943, loss: 0.5249 Iter 17944, loss: 0.4995 Iter 17945, loss: 0.4634 Iter 17946, loss: 0.4898 Iter 17947, loss: 0.4678 Iter 17948, loss: 0.4845 Iter 17949, loss: 0.4746 Iter 17950, loss: 0.4644 Iter 17951, loss: 0.4662 Iter 17952, loss: 0.4730 Iter 17953, loss: 0.4727 Iter 17954, loss: 0.4906 Iter 17955, loss: 0.4546 Iter 17956, loss: 0.5163 Iter 17957, loss: 0.4875 Iter 17958, loss: 0.4631 Iter 17959, loss: 0.4939 Iter 17960, loss: 0.4728 Iter 17961, loss: 0.4719 Iter 17962, loss: 0.5032 Iter 17963, loss: 0.4707 Iter 17964, loss: 0.4967 Iter 17965, loss: 0.4913 Iter 17966, loss: 0.5002 Iter 17967, loss: 0.4856 Iter 17968, loss: 0.4782 Iter 17969, loss: 0.4978 Iter 17970, loss: 0.4900 Iter 17971, loss: 0.4373 Iter 17972, loss: 0.4698 Iter 17973, loss: 0.4633 Iter 17974, loss: 0.4761 Iter 17975, loss: 0.4761

Iter 17976, loss: 0.4823 Iter 17977, loss: 0.4619 Iter 17978, loss: 0.4882 Iter 17979, loss: 0.4941 Iter 17980, loss: 0.4548 Iter 17981, loss: 0.4729 Iter 17982, loss: 0.4864 Iter 17983, loss: 0.4945 Iter 17984, loss: 0.4935 Iter 17985, loss: 0.4601 Iter 17986, loss: 0.5019 Iter 17987, loss: 0.4710 Iter 17988, loss: 0.5148 Iter 17989, loss: 0.5009 Iter 17990, loss: 0.4981 Iter 17991, loss: 0.4989 Iter 17992, loss: 0.4490 Iter 17993, loss: 0.4590 Iter 17994, loss: 0.4992 Iter 17995, loss: 0.4700 Iter 17996, loss: 0.4764 Iter 17997, loss: 0.4763 Iter 17998, loss: 0.4802 Iter 17999, loss: 0.4630 Iter 18000, loss: 0.4787 Iter 18001, loss: 0.4898 Iter 18002, loss: 0.4861 Iter 18003, loss: 0.5071 Iter 18004, loss: 0.5145 Iter 18005, loss: 0.4958 Iter 18006, loss: 0.4613 Iter 18007, loss: 0.5131 Iter 18008, loss: 0.4948 Iter 18009, loss: 0.4979 Iter 18010, loss: 0.4152 Iter 18011, loss: 0.4594 Iter 18012, loss: 0.5126 Iter 18013, loss: 0.4899 Iter 18014, loss: 0.4835 Iter 18015, loss: 0.4870 Iter 18016, loss: 0.4703 Iter 18017, loss: 0.4872

Iter 18018, loss: 0.4959 Iter 18019, loss: 0.4882 Iter 18020, loss: 0.4614 Iter 18021, loss: 0.5191 Iter 18022, loss: 0.4876 Iter 18023, loss: 0.5151 Iter 18024, loss: 0.4574 Iter 18025, loss: 0.4693 Iter 18026, loss: 0.4763 Iter 18027, loss: 0.4858 Iter 18028, loss: 0.4608 Iter 18029, loss: 0.4892 Iter 18030, loss: 0.4819 Iter 18031, loss: 0.4817 Iter 18032, loss: 0.5021 Iter 18033, loss: 0.4565 Iter 18034, loss: 0.4972 Iter 18035, loss: 0.4433 Iter 18036, loss: 0.4759 Iter 18037, loss: 0.4807 Iter 18038, loss: 0.4975 Iter 18039, loss: 0.4615 Iter 18040, loss: 0.4712 Iter 18041, loss: 0.4763 Iter 18042, loss: 0.5101 Iter 18043, loss: 0.4836 Iter 18044, loss: 0.4920 Iter 18045, loss: 0.4671 Iter 18046, loss: 0.4406 Iter 18047, loss: 0.4866 Iter 18048, loss: 0.4626 Iter 18049, loss: 0.4963 Iter 18050, loss: 0.5156 Iter 18051, loss: 0.4758 Iter 18052, loss: 0.4881 Iter 18053, loss: 0.5014 Iter 18054, loss: 0.4494 Iter 18055, loss: 0.4740 Iter 18056, loss: 0.4665 Iter 18057, loss: 0.4815 Iter 18058, loss: 0.4664 Iter 18059, loss: 0.4481

Iter 18060, loss: 0.4719 Iter 18061, loss: 0.5025 Iter 18062, loss: 0.4671 Iter 18063, loss: 0.4807 Iter 18064, loss: 0.4625 Iter 18065, loss: 0.4933 Iter 18066, loss: 0.4923 Iter 18067, loss: 0.4638 Iter 18068, loss: 0.4600 Iter 18069, loss: 0.5090 Iter 18070, loss: 0.4805 Iter 18071, loss: 0.4998 Iter 18072, loss: 0.4646 Iter 18073, loss: 0.4744 Iter 18074, loss: 0.4972 Iter 18075, loss: 0.4971 Iter 18076, loss: 0.4799 Iter 18077, loss: 0.4808 Iter 18078, loss: 0.5229 Iter 18079, loss: 0.4785 Iter 18080, loss: 0.4735 Iter 18081, loss: 0.4943 Iter 18082, loss: 0.4317 Iter 18083, loss: 0.4540 Iter 18084, loss: 0.5124 Iter 18085, loss: 0.4328 Iter 18086, loss: 0.4773 Iter 18087, loss: 0.4636 Iter 18088, loss: 0.5086 Iter 18089, loss: 0.4955 Iter 18090, loss: 0.4632 Iter 18091, loss: 0.4800 Iter 18092, loss: 0.4538 Iter 18093, loss: 0.4636 Iter 18094, loss: 0.5008 Iter 18095, loss: 0.5333 Iter 18096, loss: 0.4668 Iter 18097, loss: 0.5009 Iter 18098, loss: 0.4395 Iter 18099, loss: 0.4892 Iter 18100, loss: 0.4902 Iter 18101, loss: 0.4993

Iter 18102, loss: 0.4570 Iter 18103, loss: 0.4391 Iter 18104, loss: 0.4763 Iter 18105, loss: 0.4604 Iter 18106, loss: 0.4712 Iter 18107, loss: 0.4967 Iter 18108, loss: 0.4475 Iter 18109, loss: 0.4580 Iter 18110, loss: 0.4526 Iter 18111, loss: 0.4989 Iter 18112, loss: 0.4886 Iter 18113, loss: 0.4656 Iter 18114, loss: 0.4298 Iter 18115, loss: 0.4930 Iter 18116, loss: 0.4798 Iter 18117, loss: 0.4489 Iter 18118, loss: 0.4483 Iter 18119, loss: 0.4730 Iter 18120, loss: 0.5221 Iter 18121, loss: 0.4711 Iter 18122, loss: 0.4501 Iter 18123, loss: 0.4632 Iter 18124, loss: 0.4523 Iter 18125, loss: 0.4707 Iter 18126, loss: 0.4253 Iter 18127, loss: 0.4839 Iter 18128, loss: 0.4672 Iter 18129, loss: 0.4526 Iter 18130, loss: 0.4175 Iter 18131, loss: 0.5013 Iter 18132, loss: 0.4310 Iter 18133, loss: 0.4150 Iter 18134, loss: 0.4743 Iter 18135, loss: 0.4092 Iter 18136, loss: 0.4984 Iter 18137, loss: 0.5155 Iter 18138, loss: 0.4816 Iter 18139, loss: 0.5086 Iter 18140, loss: 0.4732 Iter 18141, loss: 0.4590 Iter 18142, loss: 0.4800 Iter 18143, loss: 0.4962

Iter 18144, loss: 0.4586 Iter 18145, loss: 0.4571 Iter 18146, loss: 0.4550 Iter 18147, loss: 0.4980 Iter 18148, loss: 0.5134 Iter 18149, loss: 0.4065 Iter 18150, loss: 0.4860 Iter 18151, loss: 0.4827 Iter 18152, loss: 0.4726 Iter 18153, loss: 0.4751 Iter 18154, loss: 0.4907 Iter 18155, loss: 0.4382 Iter 18156, loss: 0.4769 Iter 18157, loss: 0.5179 Iter 18158, loss: 0.4524 Iter 18159, loss: 0.4843 Iter 18160, loss: 0.4696 Iter 18161, loss: 0.4765 Iter 18162, loss: 0.4782 Iter 18163, loss: 0.4653 Iter 18164, loss: 0.5085 Iter 18165, loss: 0.4275 Iter 18166, loss: 0.4481 Iter 18167, loss: 0.4874 Iter 18168, loss: 0.5273 Iter 18169, loss: 0.4975 Iter 18170, loss: 0.5018 Iter 18171, loss: 0.4517 Iter 18172, loss: 0.4697 Iter 18173, loss: 0.4871 Iter 18174, loss: 0.4760 Iter 18175, loss: 0.4702 Iter 18176, loss: 0.4745 Iter 18177, loss: 0.4998 Iter 18178, loss: 0.4794 Iter 18179, loss: 0.4824 Iter 18180, loss: 0.4797 Iter 18181, loss: 0.4905 Iter 18182, loss: 0.4859 Iter 18183, loss: 0.4632 Iter 18184, loss: 0.4768 Iter 18185, loss: 0.4814

Iter 18186, loss: 0.4945 Iter 18187, loss: 0.4985 Iter 18188, loss: 0.4697 Iter 18189, loss: 0.4812 Iter 18190, loss: 0.5151 Iter 18191, loss: 0.4523 Iter 18192, loss: 0.4873 Iter 18193, loss: 0.5163 Iter 18194, loss: 0.4763 Iter 18195, loss: 0.4674 Iter 18196, loss: 0.4498 Iter 18197, loss: 0.5153 Iter 18198, loss: 0.4607 Iter 18199, loss: 0.4952 Iter 18200, loss: 0.4677 Iter 18201, loss: 0.4845 Iter 18202, loss: 0.5058 Iter 18203, loss: 0.4651 Iter 18204, loss: 0.4584 Iter 18205, loss: 0.4467 Iter 18206, loss: 0.4865 Iter 18207, loss: 0.5036 Iter 18208, loss: 0.4694 Iter 18209, loss: 0.4600 Iter 18210, loss: 0.4784 Iter 18211, loss: 0.4919 Iter 18212, loss: 0.4739 Iter 18213, loss: 0.4524 Iter 18214, loss: 0.4774 Iter 18215, loss: 0.4774 Iter 18216, loss: 0.4627 Iter 18217, loss: 0.4936 Iter 18218, loss: 0.5081 Iter 18219, loss: 0.4619 Iter 18220, loss: 0.4911 Iter 18221, loss: 0.4770 Iter 18222, loss: 0.4776 Iter 18223, loss: 0.4703 Iter 18224, loss: 0.4747 Iter 18225, loss: 0.4980 Iter 18226, loss: 0.4910 Iter 18227, loss: 0.4680

Iter 18228, loss: 0.4932 Iter 18229, loss: 0.4730 Iter 18230, loss: 0.4730 Iter 18231, loss: 0.4656 Iter 18232, loss: 0.4665 Iter 18233, loss: 0.4556 Iter 18234, loss: 0.4872 Iter 18235, loss: 0.4870 Iter 18236, loss: 0.4687 Iter 18237, loss: 0.4429 Iter 18238, loss: 0.4917 Iter 18239, loss: 0.4960 Iter 18240, loss: 0.4767 Iter 18241, loss: 0.4908 Iter 18242, loss: 0.4699 Iter 18243, loss: 0.4698 Iter 18244, loss: 0.4773 Iter 18245, loss: 0.4642 Iter 18246, loss: 0.4849 Iter 18247, loss: 0.4125 Iter 18248, loss: 0.4573 Iter 18249, loss: 0.4866 Iter 18250, loss: 0.5215 Iter 18251, loss: 0.5010 Iter 18252, loss: 0.4709 Iter 18253, loss: 0.4659 Iter 18254, loss: 0.5078 Iter 18255, loss: 0.5277 Iter 18256, loss: 0.4843 Iter 18257, loss: 0.4598 Iter 18258, loss: 0.4543 Iter 18259, loss: 0.4501 Iter 18260, loss: 0.4944 Iter 18261, loss: 0.4974 Iter 18262, loss: 0.4907 Iter 18263, loss: 0.4402 Iter 18264, loss: 0.5075 Iter 18265, loss: 0.5053 Iter 18266, loss: 0.4857 Iter 18267, loss: 0.5134 Iter 18268, loss: 0.5134 Iter 18269, loss: 0.4824

Iter 18270, loss: 0.4581 Iter 18271, loss: 0.4668 Iter 18272, loss: 0.4899 Iter 18273, loss: 0.5040 Iter 18274, loss: 0.4368 Iter 18275, loss: 0.4598 Iter 18276, loss: 0.4810 Iter 18277, loss: 0.5032 Iter 18278, loss: 0.4887 Iter 18279, loss: 0.4866 Iter 18280, loss: 0.4399 Iter 18281, loss: 0.4609 Iter 18282, loss: 0.5135 Iter 18283, loss: 0.4606 Iter 18284, loss: 0.4630 Iter 18285, loss: 0.4859 Iter 18286, loss: 0.4386 Iter 18287, loss: 0.5098 Iter 18288, loss: 0.4609 Iter 18289, loss: 0.4645 Iter 18290, loss: 0.4689 Iter 18291, loss: 0.5165 Iter 18292, loss: 0.4453 Iter 18293, loss: 0.4612 Iter 18294, loss: 0.4835 Iter 18295, loss: 0.4866 Iter 18296, loss: 0.4748 Iter 18297, loss: 0.4572 Iter 18298, loss: 0.4667 Iter 18299, loss: 0.4420 Iter 18300, loss: 0.4974 Iter 18301, loss: 0.4552 Iter 18302, loss: 0.4807 Iter 18303, loss: 0.4482 Iter 18304, loss: 0.4734 Iter 18305, loss: 0.4712 Iter 18306, loss: 0.4533 Iter 18307, loss: 0.4147 Iter 18308, loss: 0.4650 Iter 18309, loss: 0.4631 Iter 18310, loss: 0.4500 Iter 18311, loss: 0.4814

Iter 18312, loss: 0.4626 Iter 18313, loss: 0.4648 Iter 18314, loss: 0.4431 Iter 18315, loss: 0.4723 Iter 18316, loss: 0.4969 Iter 18317, loss: 0.4874 Iter 18318, loss: 0.4324 Iter 18319, loss: 0.4504 Iter 18320, loss: 0.4985 Iter 18321, loss: 0.4870 Iter 18322, loss: 0.4851 Iter 18323, loss: 0.4615 Iter 18324, loss: 0.4755 Iter 18325, loss: 0.4695 Iter 18326, loss: 0.4933 Iter 18327, loss: 0.4896 Iter 18328, loss: 0.4678 Iter 18329, loss: 0.4958 Iter 18330, loss: 0.5187 Iter 18331, loss: 0.4782 Iter 18332, loss: 0.4981 Iter 18333, loss: 0.4868 Iter 18334, loss: 0.4753 Iter 18335, loss: 0.4599 Iter 18336, loss: 0.4626 Iter 18337, loss: 0.4820 Iter 18338, loss: 0.4892 Iter 18339, loss: 0.4592 Iter 18340, loss: 0.4800 Iter 18341, loss: 0.4450 Iter 18342, loss: 0.4946 Iter 18343, loss: 0.4447 Iter 18344, loss: 0.4620 Iter 18345, loss: 0.4525 Iter 18346, loss: 0.4935 Iter 18347, loss: 0.4647 Iter 18348, loss: 0.4756 Iter 18349, loss: 0.4883 Iter 18350, loss: 0.5019 Iter 18351, loss: 0.4541 Iter 18352, loss: 0.4307 Iter 18353, loss: 0.5082

Iter 18354, loss: 0.4687 Iter 18355, loss: 0.4672 Iter 18356, loss: 0.4389 Iter 18357, loss: 0.4448 Iter 18358, loss: 0.5102 Iter 18359, loss: 0.4508 Iter 18360, loss: 0.4934 Iter 18361, loss: 0.4765 Iter 18362, loss: 0.4765 Iter 18363, loss: 0.4755 Iter 18364, loss: 0.4800 Iter 18365, loss: 0.4860 Iter 18366, loss: 0.4970 Iter 18367, loss: 0.4562 Iter 18368, loss: 0.4697 Iter 18369, loss: 0.4683 Iter 18370, loss: 0.4479 Iter 18371, loss: 0.4291 Iter 18372, loss: 0.4695 Iter 18373, loss: 0.4972 Iter 18374, loss: 0.4308 Iter 18375, loss: 0.4775 Iter 18376, loss: 0.4913 Iter 18377, loss: 0.4834 Iter 18378, loss: 0.4299 Iter 18379, loss: 0.4546 Iter 18380, loss: 0.5285 Iter 18381, loss: 0.4665 Iter 18382, loss: 0.4660 Iter 18383, loss: 0.4689 Iter 18384, loss: 0.5332 Iter 18385, loss: 0.5190 Iter 18386, loss: 0.5009 Iter 18387, loss: 0.4892 Iter 18388, loss: 0.4544 Iter 18389, loss: 0.4568 Iter 18390, loss: 0.4789 Iter 18391, loss: 0.5198 Iter 18392, loss: 0.4715 Iter 18393, loss: 0.4328 Iter 18394, loss: 0.4573 Iter 18395, loss: 0.4917

Iter 18396, loss: 0.5015 Iter 18397, loss: 0.4377 Iter 18398, loss: 0.4618 Iter 18399, loss: 0.4784 Iter 18400, loss: 0.4974 Iter 18401, loss: 0.4966 Iter 18402, loss: 0.4710 Iter 18403, loss: 0.4745 Iter 18404, loss: 0.4440 Iter 18405, loss: 0.5013 Iter 18406, loss: 0.4738 Iter 18407, loss: 0.4638 Iter 18408, loss: 0.4582 Iter 18409, loss: 0.4727 Iter 18410, loss: 0.4538 Iter 18411, loss: 0.4615 Iter 18412, loss: 0.4569 Iter 18413, loss: 0.4974 Iter 18414, loss: 0.4625 Iter 18415, loss: 0.4628 Iter 18416, loss: 0.5179 Iter 18417, loss: 0.4717 Iter 18418, loss: 0.4752 Iter 18419, loss: 0.4780 Iter 18420, loss: 0.4689 Iter 18421, loss: 0.4714 Iter 18422, loss: 0.4582 Iter 18423, loss: 0.4818 Iter 18424, loss: 0.4541 Iter 18425, loss: 0.4809 Iter 18426, loss: 0.4448 Iter 18427, loss: 0.4528 Iter 18428, loss: 0.5016 Iter 18429, loss: 0.4789 Iter 18430, loss: 0.4739 Iter 18431, loss: 0.5126 Iter 18432, loss: 0.4677 Iter 18433, loss: 0.4381 Iter 18434, loss: 0.4715 Iter 18435, loss: 0.5051 Iter 18436, loss: 0.4398 Iter 18437, loss: 0.4684

Iter 18438, loss: 0.4694 Iter 18439, loss: 0.4684 Iter 18440, loss: 0.4647 Iter 18441, loss: 0.4746 Iter 18442, loss: 0.4888 Iter 18443, loss: 0.4643 Iter 18444, loss: 0.5072 Iter 18445, loss: 0.5105 Iter 18446, loss: 0.4468 Iter 18447, loss: 0.4753 Iter 18448, loss: 0.4899 Iter 18449, loss: 0.4738 Iter 18450, loss: 0.4792 Iter 18451, loss: 0.4447 Iter 18452, loss: 0.4713 Iter 18453, loss: 0.4570 Iter 18454, loss: 0.5047 Iter 18455, loss: 0.4707 Iter 18456, loss: 0.4912 Iter 18457, loss: 0.4751 Iter 18458, loss: 0.4845 Iter 18459, loss: 0.4551 Iter 18460, loss: 0.5128 Iter 18461, loss: 0.4914 Iter 18462, loss: 0.4594 Iter 18463, loss: 0.4453 Iter 18464, loss: 0.4993 Iter 18465, loss: 0.4358 Iter 18466, loss: 0.5213 Iter 18467, loss: 0.4817 Iter 18468, loss: 0.4806 Iter 18469, loss: 0.4991 Iter 18470, loss: 0.5088 Iter 18471, loss: 0.4731 Iter 18472, loss: 0.4877 Iter 18473, loss: 0.4723 Iter 18474, loss: 0.4941 Iter 18475, loss: 0.4697 Iter 18476, loss: 0.4333 Iter 18477, loss: 0.4763 Iter 18478, loss: 0.4856 Iter 18479, loss: 0.5092

Iter 18480, loss: 0.4755 Iter 18481, loss: 0.4811 Iter 18482, loss: 0.4515 Iter 18483, loss: 0.4853 Iter 18484, loss: 0.5136 Iter 18485, loss: 0.5203 Iter 18486, loss: 0.5008 Iter 18487, loss: 0.5118 Iter 18488, loss: 0.4648 Iter 18489, loss: 0.4943 Iter 18490, loss: 0.5002 Iter 18491, loss: 0.4794 Iter 18492, loss: 0.4889 Iter 18493, loss: 0.4826 Iter 18494, loss: 0.4809 Iter 18495, loss: 0.4550 Iter 18496, loss: 0.4725 Iter 18497, loss: 0.4977 Iter 18498, loss: 0.5019 Iter 18499, loss: 0.5118 Iter 18500, loss: 0.4834 Iter 18501, loss: 0.4980 Iter 18502, loss: 0.5046 Iter 18503, loss: 0.4817 Iter 18504, loss: 0.4265 Iter 18505, loss: 0.4642 Iter 18506, loss: 0.5075 Iter 18507, loss: 0.4501 Iter 18508, loss: 0.4759 Iter 18509, loss: 0.4770 Iter 18510, loss: 0.5008 Iter 18511, loss: 0.4449 Iter 18512, loss: 0.4643 Iter 18513, loss: 0.5482 Iter 18514, loss: 0.5119 Iter 18515, loss: 0.4722 Iter 18516, loss: 0.5173 Iter 18517, loss: 0.4307 Iter 18518, loss: 0.4763 Iter 18519, loss: 0.4562 Iter 18520, loss: 0.4498 Iter 18521, loss: 0.4546

Iter 18522, loss: 0.4489 Iter 18523, loss: 0.4747 Iter 18524, loss: 0.5057 Iter 18525, loss: 0.4698 Iter 18526, loss: 0.4529 Iter 18527, loss: 0.4702 Iter 18528, loss: 0.4922 Iter 18529, loss: 0.4943 Iter 18530, loss: 0.4646 Iter 18531, loss: 0.4586 Iter 18532, loss: 0.4945 Iter 18533, loss: 0.4541 Iter 18534, loss: 0.4872 Iter 18535, loss: 0.4961 Iter 18536, loss: 0.4353 Iter 18537, loss: 0.4997 Iter 18538, loss: 0.4454 Iter 18539, loss: 0.4408 Iter 18540, loss: 0.4874 Iter 18541, loss: 0.4422 Iter 18542, loss: 0.4883 Iter 18543, loss: 0.4675 Iter 18544, loss: 0.4516 Iter 18545, loss: 0.4906 Iter 18546, loss: 0.4939 Iter 18547, loss: 0.4331 Iter 18548, loss: 0.4695 Iter 18549, loss: 0.5139 Iter 18550, loss: 0.4744 Iter 18551, loss: 0.4968 Iter 18552, loss: 0.5154 Iter 18553, loss: 0.4925 Iter 18554, loss: 0.5017 Iter 18555, loss: 0.4713 Iter 18556, loss: 0.4705 Iter 18557, loss: 0.4949 Iter 18558, loss: 0.5103 Iter 18559, loss: 0.4807 Iter 18560, loss: 0.4834 Iter 18561, loss: 0.5050 Iter 18562, loss: 0.4821 Iter 18563, loss: 0.4872

Iter 18564, loss: 0.4875 Iter 18565, loss: 0.4995 Iter 18566, loss: 0.4560 Iter 18567, loss: 0.4705 Iter 18568, loss: 0.4545 Iter 18569, loss: 0.4737 Iter 18570, loss: 0.5064 Iter 18571, loss: 0.4823 Iter 18572, loss: 0.4713 Iter 18573, loss: 0.5213 Iter 18574, loss: 0.4744 Iter 18575, loss: 0.4719 Iter 18576, loss: 0.4916 Iter 18577, loss: 0.4511 Iter 18578, loss: 0.4577 Iter 18579, loss: 0.5044 Iter 18580, loss: 0.4765 Iter 18581, loss: 0.5026 Iter 18582, loss: 0.4834 Iter 18583, loss: 0.4703 Iter 18584, loss: 0.4663 Iter 18585, loss: 0.4264 Iter 18586, loss: 0.4209 Iter 18587, loss: 0.4995 Iter 18588, loss: 0.4690 Iter 18589, loss: 0.4822 Iter 18590, loss: 0.4724 Iter 18591, loss: 0.4666 Iter 18592, loss: 0.4876 Iter 18593, loss: 0.4815 Iter 18594, loss: 0.4825 Iter 18595, loss: 0.4560 Iter 18596, loss: 0.4811 Iter 18597, loss: 0.4901 Iter 18598, loss: 0.4830 Iter 18599, loss: 0.4444 Iter 18600, loss: 0.4466 Iter 18601, loss: 0.4638 Iter 18602, loss: 0.4501 Iter 18603, loss: 0.4652 Iter 18604, loss: 0.4596 Iter 18605, loss: 0.4912

Iter 18606, loss: 0.4613 Iter 18607, loss: 0.5084 Iter 18608, loss: 0.4701 Iter 18609, loss: 0.4998 Iter 18610, loss: 0.4805 Iter 18611, loss: 0.4879 Iter 18612, loss: 0.4846 Iter 18613, loss: 0.4983 Iter 18614, loss: 0.4592 Iter 18615, loss: 0.4865 Iter 18616, loss: 0.4531 Iter 18617, loss: 0.4734 Iter 18618, loss: 0.4346 Iter 18619, loss: 0.4954 Iter 18620, loss: 0.4725 Iter 18621, loss: 0.4849 Iter 18622, loss: 0.4482 Iter 18623, loss: 0.4647 Iter 18624, loss: 0.4441 Iter 18625, loss: 0.4732 Iter 18626, loss: 0.4929 Iter 18627, loss: 0.4776 Iter 18628, loss: 0.4669 Iter 18629, loss: 0.4587 Iter 18630, loss: 0.4650 Iter 18631, loss: 0.4601 Iter 18632, loss: 0.4851 Iter 18633, loss: 0.5324 Iter 18634, loss: 0.4746 Iter 18635, loss: 0.4988 Iter 18636, loss: 0.4618 Iter 18637, loss: 0.4646 Iter 18638, loss: 0.4945 Iter 18639, loss: 0.4735 Iter 18640, loss: 0.5105 Iter 18641, loss: 0.4923 Iter 18642, loss: 0.4739 Iter 18643, loss: 0.4602 Iter 18644, loss: 0.4401 Iter 18645, loss: 0.4683 Iter 18646, loss: 0.4453 Iter 18647, loss: 0.4619

Iter 18648, loss: 0.5152 Iter 18649, loss: 0.4590 Iter 18650, loss: 0.4720 Iter 18651, loss: 0.4674 Iter 18652, loss: 0.4717 Iter 18653, loss: 0.5059 Iter 18654, loss: 0.4703 Iter 18655, loss: 0.4598 Iter 18656, loss: 0.4605 Iter 18657, loss: 0.4770 Iter 18658, loss: 0.4989 Iter 18659, loss: 0.4474 Iter 18660, loss: 0.4699 Iter 18661, loss: 0.4972 Iter 18662, loss: 0.4876 Iter 18663, loss: 0.4781 Iter 18664, loss: 0.4513 Iter 18665, loss: 0.4280 Iter 18666, loss: 0.4936 Iter 18667, loss: 0.4694 Iter 18668, loss: 0.4616 Iter 18669, loss: 0.4853 Iter 18670, loss: 0.4587 Iter 18671, loss: 0.5072 Iter 18672, loss: 0.4825 Iter 18673, loss: 0.4686 Iter 18674, loss: 0.5049 Iter 18675, loss: 0.5046 Iter 18676, loss: 0.4549 Iter 18677, loss: 0.5103 Iter 18678, loss: 0.4765 Iter 18679, loss: 0.4803 Iter 18680, loss: 0.4962 Iter 18681, loss: 0.4614 Iter 18682, loss: 0.4253 Iter 18683, loss: 0.4916 Iter 18684, loss: 0.5129 Iter 18685, loss: 0.4363 Iter 18686, loss: 0.4875 Iter 18687, loss: 0.4704 Iter 18688, loss: 0.4437 Iter 18689, loss: 0.4893

Iter 18690, loss: 0.4759 Iter 18691, loss: 0.4891 Iter 18692, loss: 0.4932 Iter 18693, loss: 0.4895 Iter 18694, loss: 0.5138 Iter 18695, loss: 0.4731 Iter 18696, loss: 0.4520 Iter 18697, loss: 0.4664 Iter 18698, loss: 0.4380 Iter 18699, loss: 0.4588 Iter 18700, loss: 0.4617 Iter 18701, loss: 0.4736 Iter 18702, loss: 0.4838 Iter 18703, loss: 0.4812 Iter 18704, loss: 0.4599 Iter 18705, loss: 0.4654 Iter 18706, loss: 0.4806 Iter 18707, loss: 0.5209 Iter 18708, loss: 0.4699 Iter 18709, loss: 0.4680 Iter 18710, loss: 0.4909 Iter 18711, loss: 0.4961 Iter 18712, loss: 0.5047 Iter 18713, loss: 0.4462 Iter 18714, loss: 0.4529 Iter 18715, loss: 0.5038 Iter 18716, loss: 0.4529 Iter 18717, loss: 0.4792 Iter 18718, loss: 0.4739 Iter 18719, loss: 0.4701 Iter 18720, loss: 0.4656 Iter 18721, loss: 0.4723 Iter 18722, loss: 0.4991 Iter 18723, loss: 0.4446 Iter 18724, loss: 0.4824 Iter 18725, loss: 0.4931 Iter 18726, loss: 0.4710 Iter 18727, loss: 0.4552 Iter 18728, loss: 0.4405 Iter 18729, loss: 0.5074 Iter 18730, loss: 0.4728 Iter 18731, loss: 0.4787

Iter 18732, loss: 0.4484 Iter 18733, loss: 0.5025 Iter 18734, loss: 0.4803 Iter 18735, loss: 0.4868 Iter 18736, loss: 0.4728 Iter 18737, loss: 0.4363 Iter 18738, loss: 0.4827 Iter 18739, loss: 0.4487 Iter 18740, loss: 0.4577 Iter 18741, loss: 0.5017 Iter 18742, loss: 0.4850 Iter 18743, loss: 0.4824 Iter 18744, loss: 0.4935 Iter 18745, loss: 0.4713 Iter 18746, loss: 0.4653 Iter 18747, loss: 0.4720 Iter 18748, loss: 0.4630 Iter 18749, loss: 0.5089 Iter 18750, loss: 0.4670 Iter 18751, loss: 0.5071 Iter 18752, loss: 0.4871 Iter 18753, loss: 0.4692 Iter 18754, loss: 0.4449 Iter 18755, loss: 0.4191 Iter 18756, loss: 0.4956 Iter 18757, loss: 0.4852 Iter 18758, loss: 0.4698 Iter 18759, loss: 0.4820 Iter 18760, loss: 0.4722 Iter 18761, loss: 0.5065 Iter 18762, loss: 0.4514 Iter 18763, loss: 0.4806 Iter 18764, loss: 0.4904 Iter 18765, loss: 0.4897 Iter 18766, loss: 0.4832 Iter 18767, loss: 0.5030 Iter 18768, loss: 0.4696 Iter 18769, loss: 0.4901 Iter 18770, loss: 0.4974 Iter 18771, loss: 0.4844 Iter 18772, loss: 0.4489 Iter 18773, loss: 0.4595

Iter 18774, loss: 0.4894 Iter 18775, loss: 0.4478 Iter 18776, loss: 0.4955 Iter 18777, loss: 0.4691 Iter 18778, loss: 0.5191 Iter 18779, loss: 0.4856 Iter 18780, loss: 0.4859 Iter 18781, loss: 0.4695 Iter 18782, loss: 0.4736 Iter 18783, loss: 0.4846 Iter 18784, loss: 0.4903 Iter 18785, loss: 0.4647 Iter 18786, loss: 0.4866 Iter 18787, loss: 0.4986 Iter 18788, loss: 0.4854 Iter 18789, loss: 0.4886 Iter 18790, loss: 0.4778 Iter 18791, loss: 0.4942 Iter 18792, loss: 0.4785 Iter 18793, loss: 0.4518 Iter 18794, loss: 0.5051 Iter 18795, loss: 0.5129 Iter 18796, loss: 0.5109 Iter 18797, loss: 0.4673 Iter 18798, loss: 0.4652 Iter 18799, loss: 0.4511 Iter 18800, loss: 0.4993 Iter 18801, loss: 0.4528 Iter 18802, loss: 0.4640 Iter 18803, loss: 0.4579 Iter 18804, loss: 0.4697 Iter 18805, loss: 0.4547 Iter 18806, loss: 0.4643 Iter 18807, loss: 0.5299 Iter 18808, loss: 0.4555 Iter 18809, loss: 0.4827 Iter 18810, loss: 0.4737 Iter 18811, loss: 0.4912 Iter 18812, loss: 0.4975 Iter 18813, loss: 0.4627 Iter 18814, loss: 0.4283 Iter 18815, loss: 0.4923

Iter 18816, loss: 0.4759 Iter 18817, loss: 0.5045 Iter 18818, loss: 0.4327 Iter 18819, loss: 0.4678 Iter 18820, loss: 0.4553 Iter 18821, loss: 0.4364 Iter 18822, loss: 0.4808 Iter 18823, loss: 0.4375 Iter 18824, loss: 0.4704 Iter 18825, loss: 0.4486 Iter 18826, loss: 0.4917 Iter 18827, loss: 0.4602 Iter 18828, loss: 0.5129 Iter 18829, loss: 0.4584 Iter 18830, loss: 0.4651 Iter 18831, loss: 0.4523 Iter 18832, loss: 0.4509 Iter 18833, loss: 0.4900 Iter 18834, loss: 0.4736 Iter 18835, loss: 0.5006 Iter 18836, loss: 0.4491 Iter 18837, loss: 0.4699 Iter 18838, loss: 0.4495 Iter 18839, loss: 0.4757 Iter 18840, loss: 0.4544 Iter 18841, loss: 0.5287 Iter 18842, loss: 0.4746 Iter 18843, loss: 0.4648 Iter 18844, loss: 0.4696 Iter 18845, loss: 0.5047 Iter 18846, loss: 0.4899 Iter 18847, loss: 0.4154 Iter 18848, loss: 0.4818 Iter 18849, loss: 0.5278 Iter 18850, loss: 0.4675 Iter 18851, loss: 0.4689 Iter 18852, loss: 0.4433 Iter 18853, loss: 0.5072 Iter 18854, loss: 0.4331 Iter 18855, loss: 0.4689 Iter 18856, loss: 0.5386 Iter 18857, loss: 0.4815

Iter 18858, loss: 0.4320 Iter 18859, loss: 0.4726 Iter 18860, loss: 0.5271 Iter 18861, loss: 0.4268 Iter 18862, loss: 0.4571 Iter 18863, loss: 0.4976 Iter 18864, loss: 0.5071 Iter 18865, loss: 0.4813 Iter 18866, loss: 0.4542 Iter 18867, loss: 0.4605 Iter 18868, loss: 0.4610 Iter 18869, loss: 0.4749 Iter 18870, loss: 0.5000 Iter 18871, loss: 0.4713 Iter 18872, loss: 0.4937 Iter 18873, loss: 0.4783 Iter 18874, loss: 0.4660 Iter 18875, loss: 0.4769 Iter 18876, loss: 0.4770 Iter 18877, loss: 0.4567 Iter 18878, loss: 0.4391 Iter 18879, loss: 0.4570 Iter 18880, loss: 0.4777 Iter 18881, loss: 0.4391 Iter 18882, loss: 0.5136 Iter 18883, loss: 0.4685 Iter 18884, loss: 0.4599 Iter 18885, loss: 0.4701 Iter 18886, loss: 0.4500 Iter 18887, loss: 0.4805 Iter 18888, loss: 0.4748 Iter 18889, loss: 0.4820 Iter 18890, loss: 0.4584 Iter 18891, loss: 0.4684 Iter 18892, loss: 0.4648 Iter 18893, loss: 0.5131 Iter 18894, loss: 0.4867 Iter 18895, loss: 0.4738 Iter 18896, loss: 0.4693 Iter 18897, loss: 0.4945 Iter 18898, loss: 0.4723 Iter 18899, loss: 0.4655

Iter 18900, loss: 0.5042 Iter 18901, loss: 0.4991 Iter 18902, loss: 0.4773 Iter 18903, loss: 0.4948 Iter 18904, loss: 0.4803 Iter 18905, loss: 0.4721 Iter 18906, loss: 0.4826 Iter 18907, loss: 0.4767 Iter 18908, loss: 0.4828 Iter 18909, loss: 0.4820 Iter 18910, loss: 0.4499 Iter 18911, loss: 0.4869 Iter 18912, loss: 0.4724 Iter 18913, loss: 0.4821 Iter 18914, loss: 0.4784 Iter 18915, loss: 0.4864 Iter 18916, loss: 0.5019 Iter 18917, loss: 0.4950 Iter 18918, loss: 0.5314 Iter 18919, loss: 0.4681 Iter 18920, loss: 0.4476 Iter 18921, loss: 0.4846 Iter 18922, loss: 0.4577 Iter 18923, loss: 0.4734 Iter 18924, loss: 0.5091 Iter 18925, loss: 0.4587 Iter 18926, loss: 0.4973 Iter 18927, loss: 0.4383 Iter 18928, loss: 0.4907 Iter 18929, loss: 0.4794 Iter 18930, loss: 0.4808 Iter 18931, loss: 0.4831 Iter 18932, loss: 0.4969 Iter 18933, loss: 0.4698 Iter 18934, loss: 0.4568 Iter 18935, loss: 0.4788 Iter 18936, loss: 0.4504 Iter 18937, loss: 0.4632 Iter 18938, loss: 0.4486 Iter 18939, loss: 0.4736 Iter 18940, loss: 0.4555 Iter 18941, loss: 0.4730

Iter 18942, loss: 0.4837 Iter 18943, loss: 0.4649 Iter 18944, loss: 0.4519 Iter 18945, loss: 0.4933 Iter 18946, loss: 0.4718 Iter 18947, loss: 0.4883 Iter 18948, loss: 0.4368 Iter 18949, loss: 0.4397 Iter 18950, loss: 0.5242 Iter 18951, loss: 0.4642 Iter 18952, loss: 0.4544 Iter 18953, loss: 0.4710 Iter 18954, loss: 0.4956 Iter 18955, loss: 0.4630 Iter 18956, loss: 0.4218 Iter 18957, loss: 0.5116 Iter 18958, loss: 0.4783 Iter 18959, loss: 0.4685 Iter 18960, loss: 0.4704 Iter 18961, loss: 0.4787 Iter 18962, loss: 0.5026 Iter 18963, loss: 0.4230 Iter 18964, loss: 0.4798 Iter 18965, loss: 0.4581 Iter 18966, loss: 0.4822 Iter 18967, loss: 0.5180 Iter 18968, loss: 0.4587 Iter 18969, loss: 0.4819 Iter 18970, loss: 0.4958 Iter 18971, loss: 0.4896 Iter 18972, loss: 0.4757 Iter 18973, loss: 0.4805 Iter 18974, loss: 0.4704 Iter 18975, loss: 0.4693 Iter 18976, loss: 0.4853 Iter 18977, loss: 0.5145 Iter 18978, loss: 0.4593 Iter 18979, loss: 0.5011 Iter 18980, loss: 0.4628 Iter 18981, loss: 0.4309 Iter 18982, loss: 0.4718 Iter 18983, loss: 0.4787

Iter 18984, loss: 0.4648 Iter 18985, loss: 0.4841 Iter 18986, loss: 0.4446 Iter 18987, loss: 0.4710 Iter 18988, loss: 0.4673 Iter 18989, loss: 0.4915 Iter 18990, loss: 0.4963 Iter 18991, loss: 0.4791 Iter 18992, loss: 0.5219 Iter 18993, loss: 0.4816 Iter 18994, loss: 0.4791 Iter 18995, loss: 0.4781 Iter 18996, loss: 0.5324 Iter 18997, loss: 0.4632 Iter 18998, loss: 0.4674 Iter 18999, loss: 0.4675 Iter 19000, loss: 0.4729 Iter 19001, loss: 0.4800 Iter 19002, loss: 0.4747 Iter 19003, loss: 0.4906 Iter 19004, loss: 0.4792 Iter 19005, loss: 0.4626 Iter 19006, loss: 0.4669 Iter 19007, loss: 0.5103 Iter 19008, loss: 0.4470 Iter 19009, loss: 0.4550 Iter 19010, loss: 0.5025 Iter 19011, loss: 0.4723 Iter 19012, loss: 0.4751 Iter 19013, loss: 0.4612 Iter 19014, loss: 0.5115 Iter 19015, loss: 0.4731 Iter 19016, loss: 0.4507 Iter 19017, loss: 0.4854 Iter 19018, loss: 0.4707 Iter 19019, loss: 0.4574 Iter 19020, loss: 0.4714 Iter 19021, loss: 0.4871 Iter 19022, loss: 0.4700 Iter 19023, loss: 0.4591 Iter 19024, loss: 0.4800 Iter 19025, loss: 0.4513

Iter 19026, loss: 0.5038 Iter 19027, loss: 0.4821 Iter 19028, loss: 0.4520 Iter 19029, loss: 0.4952 Iter 19030, loss: 0.4737 Iter 19031, loss: 0.4884 Iter 19032, loss: 0.4517 Iter 19033, loss: 0.4911 Iter 19034, loss: 0.4833 Iter 19035, loss: 0.4941 Iter 19036, loss: 0.4658 Iter 19037, loss: 0.4538 Iter 19038, loss: 0.4469 Iter 19039, loss: 0.4752 Iter 19040, loss: 0.4437 Iter 19041, loss: 0.4696 Iter 19042, loss: 0.5167 Iter 19043, loss: 0.4415 Iter 19044, loss: 0.4885 Iter 19045, loss: 0.4616 Iter 19046, loss: 0.4786 Iter 19047, loss: 0.4550 Iter 19048, loss: 0.5194 Iter 19049, loss: 0.4928 Iter 19050, loss: 0.4714 Iter 19051, loss: 0.4817 Iter 19052, loss: 0.4612 Iter 19053, loss: 0.4618 Iter 19054, loss: 0.4510 Iter 19055, loss: 0.5006 Iter 19056, loss: 0.4822 Iter 19057, loss: 0.4206 Iter 19058, loss: 0.4741 Iter 19059, loss: 0.4556 Iter 19060, loss: 0.4988 Iter 19061, loss: 0.4830 Iter 19062, loss: 0.4687 Iter 19063, loss: 0.4906 Iter 19064, loss: 0.4662 Iter 19065, loss: 0.4534 Iter 19066, loss: 0.4608 Iter 19067, loss: 0.4571

Iter 19068, loss: 0.4624 Iter 19069, loss: 0.4696 Iter 19070, loss: 0.5043 Iter 19071, loss: 0.4758 Iter 19072, loss: 0.4556 Iter 19073, loss: 0.4363 Iter 19074, loss: 0.4919 Iter 19075, loss: 0.4757 Iter 19076, loss: 0.4764 Iter 19077, loss: 0.4061 Iter 19078, loss: 0.4524 Iter 19079, loss: 0.4670 Iter 19080, loss: 0.4317 Iter 19081, loss: 0.4894 Iter 19082, loss: 0.4840 Iter 19083, loss: 0.4645 Iter 19084, loss: 0.4579 Iter 19085, loss: 0.5017 Iter 19086, loss: 0.5242 Iter 19087, loss: 0.4531 Iter 19088, loss: 0.4808 Iter 19089, loss: 0.5032 Iter 19090, loss: 0.4448 Iter 19091, loss: 0.5272 Iter 19092, loss: 0.4252 Iter 19093, loss: 0.4533 Iter 19094, loss: 0.4887 Iter 19095, loss: 0.4656 Iter 19096, loss: 0.4716 Iter 19097, loss: 0.4844 Iter 19098, loss: 0.4665 Iter 19099, loss: 0.4725 Iter 19100, loss: 0.4564 Iter 19101, loss: 0.4855 Iter 19102, loss: 0.4863 Iter 19103, loss: 0.4689 Iter 19104, loss: 0.4804 Iter 19105, loss: 0.5298 Iter 19106, loss: 0.4940 Iter 19107, loss: 0.4549 Iter 19108, loss: 0.4314 Iter 19109, loss: 0.4540

Iter 19110, loss: 0.4886 Iter 19111, loss: 0.4671 Iter 19112, loss: 0.4901 Iter 19113, loss: 0.4799 Iter 19114, loss: 0.4840 Iter 19115, loss: 0.4856 Iter 19116, loss: 0.5051 Iter 19117, loss: 0.4717 Iter 19118, loss: 0.4698 Iter 19119, loss: 0.4290 Iter 19120, loss: 0.4785 Iter 19121, loss: 0.4464 Iter 19122, loss: 0.4792 Iter 19123, loss: 0.4362 Iter 19124, loss: 0.4588 Iter 19125, loss: 0.4985 Iter 19126, loss: 0.4863 Iter 19127, loss: 0.4466 Iter 19128, loss: 0.4613 Iter 19129, loss: 0.4903 Iter 19130, loss: 0.4968 Iter 19131, loss: 0.4976 Iter 19132, loss: 0.4374 Iter 19133, loss: 0.5048 Iter 19134, loss: 0.4782 Iter 19135, loss: 0.4411 Iter 19136, loss: 0.4917 Iter 19137, loss: 0.4332 Iter 19138, loss: 0.4760 Iter 19139, loss: 0.4410 Iter 19140, loss: 0.4657 Iter 19141, loss: 0.5054 Iter 19142, loss: 0.4584 Iter 19143, loss: 0.4926 Iter 19144, loss: 0.4435 Iter 19145, loss: 0.4414 Iter 19146, loss: 0.4558 Iter 19147, loss: 0.4514 Iter 19148, loss: 0.5168 Iter 19149, loss: 0.4984 Iter 19150, loss: 0.5031 Iter 19151, loss: 0.4564

Iter 19152, loss: 0.4731 Iter 19153, loss: 0.4224 Iter 19154, loss: 0.4974 Iter 19155, loss: 0.4585 Iter 19156, loss: 0.4915 Iter 19157, loss: 0.4980 Iter 19158, loss: 0.4575 Iter 19159, loss: 0.4655 Iter 19160, loss: 0.4738 Iter 19161, loss: 0.4709 Iter 19162, loss: 0.4493 Iter 19163, loss: 0.5005 Iter 19164, loss: 0.4709 Iter 19165, loss: 0.5006 Iter 19166, loss: 0.5059 Iter 19167, loss: 0.4711 Iter 19168, loss: 0.4764 Iter 19169, loss: 0.5101 Iter 19170, loss: 0.4711 Iter 19171, loss: 0.4773 Iter 19172, loss: 0.5004 Iter 19173, loss: 0.4427 Iter 19174, loss: 0.4878 Iter 19175, loss: 0.4863 Iter 19176, loss: 0.4992 Iter 19177, loss: 0.4918 Iter 19178, loss: 0.4628 Iter 19179, loss: 0.5247 Iter 19180, loss: 0.4433 Iter 19181, loss: 0.5144 Iter 19182, loss: 0.5073 Iter 19183, loss: 0.4954 Iter 19184, loss: 0.4982 Iter 19185, loss: 0.4730 Iter 19186, loss: 0.4790 Iter 19187, loss: 0.4497 Iter 19188, loss: 0.5055 Iter 19189, loss: 0.4432 Iter 19190, loss: 0.4530 Iter 19191, loss: 0.4748 Iter 19192, loss: 0.4582 Iter 19193, loss: 0.4972

Iter 19194, loss: 0.4381 Iter 19195, loss: 0.5021 Iter 19196, loss: 0.4845 Iter 19197, loss: 0.4908 Iter 19198, loss: 0.4814 Iter 19199, loss: 0.4888 Iter 19200, loss: 0.4790 Iter 19201, loss: 0.4425 Iter 19202, loss: 0.4795 Iter 19203, loss: 0.4944 Iter 19204, loss: 0.5041 Iter 19205, loss: 0.5045 Iter 19206, loss: 0.4784 Iter 19207, loss: 0.4480 Iter 19208, loss: 0.4610 Iter 19209, loss: 0.4922 Iter 19210, loss: 0.4813 Iter 19211, loss: 0.4897 Iter 19212, loss: 0.4224 Iter 19213, loss: 0.4739 Iter 19214, loss: 0.4756 Iter 19215, loss: 0.4533 Iter 19216, loss: 0.4577 Iter 19217, loss: 0.4744 Iter 19218, loss: 0.5212 Iter 19219, loss: 0.4407 Iter 19220, loss: 0.4812 Iter 19221, loss: 0.4283 Iter 19222, loss: 0.4569 Iter 19223, loss: 0.4246 Iter 19224, loss: 0.4848 Iter 19225, loss: 0.5115 Iter 19226, loss: 0.4718 Iter 19227, loss: 0.4688 Iter 19228, loss: 0.4352 Iter 19229, loss: 0.5063 Iter 19230, loss: 0.4430 Iter 19231, loss: 0.4518 Iter 19232, loss: 0.4702 Iter 19233, loss: 0.5116 Iter 19234, loss: 0.4798 Iter 19235, loss: 0.5052

Iter 19236, loss: 0.5024 Iter 19237, loss: 0.4365 Iter 19238, loss: 0.4426 Iter 19239, loss: 0.4735 Iter 19240, loss: 0.5018 Iter 19241, loss: 0.4968 Iter 19242, loss: 0.4883 Iter 19243, loss: 0.4414 Iter 19244, loss: 0.4766 Iter 19245, loss: 0.4459 Iter 19246, loss: 0.4869 Iter 19247, loss: 0.4681 Iter 19248, loss: 0.4854 Iter 19249, loss: 0.4380 Iter 19250, loss: 0.4852 Iter 19251, loss: 0.4468 Iter 19252, loss: 0.4995 Iter 19253, loss: 0.4863 Iter 19254, loss: 0.4987 Iter 19255, loss: 0.4584 Iter 19256, loss: 0.4991 Iter 19257, loss: 0.4365 Iter 19258, loss: 0.4535 Iter 19259, loss: 0.4488 Iter 19260, loss: 0.5092 Iter 19261, loss: 0.4247 Iter 19262, loss: 0.4654 Iter 19263, loss: 0.4702 Iter 19264, loss: 0.4758 Iter 19265, loss: 0.4409 Iter 19266, loss: 0.4969 Iter 19267, loss: 0.5254 Iter 19268, loss: 0.4694 Iter 19269, loss: 0.4674 Iter 19270, loss: 0.4810 Iter 19271, loss: 0.5276 Iter 19272, loss: 0.4771 Iter 19273, loss: 0.4908 Iter 19274, loss: 0.4420 Iter 19275, loss: 0.5025 Iter 19276, loss: 0.4921 Iter 19277, loss: 0.5101

Iter 19278, loss: 0.4700 Iter 19279, loss: 0.4347 Iter 19280, loss: 0.4926 Iter 19281, loss: 0.4971 Iter 19282, loss: 0.4250 Iter 19283, loss: 0.4809 Iter 19284, loss: 0.4677 Iter 19285, loss: 0.4755 Iter 19286, loss: 0.4649 Iter 19287, loss: 0.4490 Iter 19288, loss: 0.4358 Iter 19289, loss: 0.4528 Iter 19290, loss: 0.4645 Iter 19291, loss: 0.4883 Iter 19292, loss: 0.4567 Iter 19293, loss: 0.4844 Iter 19294, loss: 0.5195 Iter 19295, loss: 0.4965 Iter 19296, loss: 0.5040 Iter 19297, loss: 0.4485 Iter 19298, loss: 0.4274 Iter 19299, loss: 0.4810 Iter 19300, loss: 0.4705 Iter 19301, loss: 0.4666 Iter 19302, loss: 0.4619 Iter 19303, loss: 0.4641 Iter 19304, loss: 0.4826 Iter 19305, loss: 0.4922 Iter 19306, loss: 0.4801 Iter 19307, loss: 0.4845 Iter 19308, loss: 0.5016 Iter 19309, loss: 0.4463 Iter 19310, loss: 0.4890 Iter 19311, loss: 0.4792 Iter 19312, loss: 0.4651 Iter 19313, loss: 0.4792 Iter 19314, loss: 0.4572 Iter 19315, loss: 0.5172 Iter 19316, loss: 0.4815 Iter 19317, loss: 0.4695 Iter 19318, loss: 0.4505 Iter 19319, loss: 0.4546

Iter 19320, loss: 0.5027 Iter 19321, loss: 0.4780 Iter 19322, loss: 0.5037 Iter 19323, loss: 0.4694 Iter 19324, loss: 0.4591 Iter 19325, loss: 0.4507 Iter 19326, loss: 0.4957 Iter 19327, loss: 0.5033 Iter 19328, loss: 0.4465 Iter 19329, loss: 0.4678 Iter 19330, loss: 0.4865 Iter 19331, loss: 0.4738 Iter 19332, loss: 0.4896 Iter 19333, loss: 0.5156 Iter 19334, loss: 0.4741 Iter 19335, loss: 0.5033 Iter 19336, loss: 0.4502 Iter 19337, loss: 0.4412 Iter 19338, loss: 0.5126 Iter 19339, loss: 0.4569 Iter 19340, loss: 0.4524 Iter 19341, loss: 0.5074 Iter 19342, loss: 0.4588 Iter 19343, loss: 0.4370 Iter 19344, loss: 0.4702 Iter 19345, loss: 0.4994 Iter 19346, loss: 0.4720 Iter 19347, loss: 0.5080 Iter 19348, loss: 0.4875 Iter 19349, loss: 0.4899 Iter 19350, loss: 0.4285 Iter 19351, loss: 0.4707 Iter 19352, loss: 0.4441 Iter 19353, loss: 0.5050 Iter 19354, loss: 0.4943 Iter 19355, loss: 0.4811 Iter 19356, loss: 0.5001 Iter 19357, loss: 0.4954 Iter 19358, loss: 0.4507 Iter 19359, loss: 0.4448 Iter 19360, loss: 0.4983 Iter 19361, loss: 0.4770

Iter 19362, loss: 0.4872 Iter 19363, loss: 0.5135 Iter 19364, loss: 0.4634 Iter 19365, loss: 0.4586 Iter 19366, loss: 0.5056 Iter 19367, loss: 0.4410 Iter 19368, loss: 0.4864 Iter 19369, loss: 0.4711 Iter 19370, loss: 0.4955 Iter 19371, loss: 0.4852 Iter 19372, loss: 0.5184 Iter 19373, loss: 0.4816 Iter 19374, loss: 0.4746 Iter 19375, loss: 0.4599 Iter 19376, loss: 0.4819 Iter 19377, loss: 0.4707 Iter 19378, loss: 0.5023 Iter 19379, loss: 0.5466 Iter 19380, loss: 0.4831 Iter 19381, loss: 0.4590 Iter 19382, loss: 0.4854 Iter 19383, loss: 0.4687 Iter 19384, loss: 0.4828 Iter 19385, loss: 0.4609 Iter 19386, loss: 0.4272 Iter 19387, loss: 0.4466 Iter 19388, loss: 0.4333 Iter 19389, loss: 0.4700 Iter 19390, loss: 0.4484 Iter 19391, loss: 0.5067 Iter 19392, loss: 0.4908 Iter 19393, loss: 0.4437 Iter 19394, loss: 0.4882 Iter 19395, loss: 0.4545 Iter 19396, loss: 0.4480 Iter 19397, loss: 0.4576 Iter 19398, loss: 0.5016 Iter 19399, loss: 0.4824 Iter 19400, loss: 0.4655 Iter 19401, loss: 0.4758 Iter 19402, loss: 0.4912

Iter 19403, loss: 0.4974 Iter 19404, loss: 0.4759 Iter 19405, loss: 0.4519 Iter 19406, loss: 0.4983 Iter 19407, loss: 0.4912 Iter 19408, loss: 0.5151 Iter 19409, loss: 0.5006 Iter 19410, loss: 0.4849 Iter 19411, loss: 0.4714 Iter 19412, loss: 0.4807 Iter 19413, loss: 0.4680 Iter 19414, loss: 0.4774 Iter 19415, loss: 0.4729 Iter 19416, loss: 0.4397 Iter 19417, loss: 0.4775 Iter 19418, loss: 0.4997 Iter 19419, loss: 0.4830 Iter 19420, loss: 0.4815 Iter 19421, loss: 0.5039 Iter 19422, loss: 0.4885 Iter 19423, loss: 0.4856 Iter 19424, loss: 0.4691 Iter 19425, loss: 0.4604 Iter 19426, loss: 0.4969 Iter 19427, loss: 0.4961 Iter 19428, loss: 0.5095 Iter 19429, loss: 0.4634 Iter 19430, loss: 0.4706 Iter 19431, loss: 0.4526 Iter 19432, loss: 0.4705 Iter 19433, loss: 0.4985 Iter 19434, loss: 0.4612 Iter 19435, loss: 0.4764 Iter 19436, loss: 0.5004 Iter 19437, loss: 0.4733 Iter 19438, loss: 0.4235 Iter 19439, loss: 0.4814 Iter 19440, loss: 0.4377 Iter 19441, loss: 0.4578 Iter 19442, loss: 0.4381 Iter 19443, loss: 0.4528 Iter 19444, loss: 0.4902

Iter 19445, loss: 0.4639 Iter 19446, loss: 0.4850 Iter 19447, loss: 0.4973 Iter 19448, loss: 0.4943 Iter 19449, loss: 0.5284 Iter 19450, loss: 0.5361 Iter 19451, loss: 0.4919 Iter 19452, loss: 0.4711 Iter 19453, loss: 0.4650 Iter 19454, loss: 0.4310 Iter 19455, loss: 0.4379 Iter 19456, loss: 0.4389 Iter 19457, loss: 0.4906 Iter 19458, loss: 0.5413 Iter 19459, loss: 0.4849 Iter 19460, loss: 0.5104 Iter 19461, loss: 0.4611 Iter 19462, loss: 0.3955 Iter 19463, loss: 0.4588 Iter 19464, loss: 0.4569 Iter 19465, loss: 0.4263 Iter 19466, loss: 0.4908 Iter 19467, loss: 0.5238 Iter 19468, loss: 0.4738 Iter 19469, loss: 0.4786 Iter 19470, loss: 0.5104 Iter 19471, loss: 0.4851 Iter 19472, loss: 0.4749 Iter 19473, loss: 0.4731 Iter 19474, loss: 0.4670 Iter 19475, loss: 0.4724 Iter 19476, loss: 0.4989 Iter 19477, loss: 0.4643 Iter 19478, loss: 0.4315 Iter 19479, loss: 0.4939 Iter 19480, loss: 0.4652 Iter 19481, loss: 0.4885 Iter 19482, loss: 0.4400 Iter 19483, loss: 0.4964 Iter 19484, loss: 0.4880 Iter 19485, loss: 0.4583 Iter 19486, loss: 0.4663

Iter 19487, loss: 0.4853 Iter 19488, loss: 0.4841 Iter 19489, loss: 0.4902 Iter 19490, loss: 0.4816 Iter 19491, loss: 0.4865 Iter 19492, loss: 0.4913 Iter 19493, loss: 0.4817 Iter 19494, loss: 0.4651 Iter 19495, loss: 0.5117 Iter 19496, loss: 0.4558 Iter 19497, loss: 0.4582 Iter 19498, loss: 0.4937 Iter 19499, loss: 0.4761 Iter 19500, loss: 0.4561 Iter 19501, loss: 0.4496 Iter 19502, loss: 0.4898 Iter 19503, loss: 0.4994 Iter 19504, loss: 0.4399 Iter 19505, loss: 0.4573 Iter 19506, loss: 0.4491 Iter 19507, loss: 0.4903 Iter 19508, loss: 0.5087 Iter 19509, loss: 0.4813 Iter 19510, loss: 0.4252 Iter 19511, loss: 0.4896 Iter 19512, loss: 0.4526 Iter 19513, loss: 0.4684 Iter 19514, loss: 0.4538 Iter 19515, loss: 0.4797 Iter 19516, loss: 0.4656 Iter 19517, loss: 0.4672 Iter 19518, loss: 0.4822 Iter 19519, loss: 0.4593 Iter 19520, loss: 0.4849 Iter 19521, loss: 0.4599 Iter 19522, loss: 0.4537 Iter 19523, loss: 0.4596 Iter 19524, loss: 0.4567 Iter 19525, loss: 0.4764 Iter 19526, loss: 0.4814 Iter 19527, loss: 0.4511 Iter 19528, loss: 0.4397

Iter 19529, loss: 0.4759 Iter 19530, loss: 0.4489 Iter 19531, loss: 0.4863 Iter 19532, loss: 0.4869 Iter 19533, loss: 0.4479 Iter 19534, loss: 0.5224 Iter 19535, loss: 0.4375 Iter 19536, loss: 0.4544 Iter 19537, loss: 0.4550 Iter 19538, loss: 0.4625 Iter 19539, loss: 0.4510 Iter 19540, loss: 0.4907 Iter 19541, loss: 0.4803 Iter 19542, loss: 0.4652 Iter 19543, loss: 0.4829 Iter 19544, loss: 0.4999 Iter 19545, loss: 0.4491 Iter 19546, loss: 0.5244 Iter 19547, loss: 0.4452 Iter 19548, loss: 0.4989 Iter 19549, loss: 0.4545 Iter 19550, loss: 0.4992 Iter 19551, loss: 0.4779 Iter 19552, loss: 0.5165 Iter 19553, loss: 0.4947 Iter 19554, loss: 0.4797 Iter 19555, loss: 0.4483 Iter 19556, loss: 0.4516 Iter 19557, loss: 0.4881 Iter 19558, loss: 0.4449 Iter 19559, loss: 0.5054 Iter 19560, loss: 0.4823 Iter 19561, loss: 0.4576 Iter 19562, loss: 0.4992 Iter 19563, loss: 0.5178 Iter 19564, loss: 0.4732 Iter 19565, loss: 0.4484 Iter 19566, loss: 0.4379 Iter 19567, loss: 0.5170 Iter 19568, loss: 0.4805 Iter 19569, loss: 0.4645 Iter 19570, loss: 0.5102

Iter 19571, loss: 0.4762 Iter 19572, loss: 0.4919 Iter 19573, loss: 0.5090 Iter 19574, loss: 0.4811 Iter 19575, loss: 0.4763 Iter 19576, loss: 0.4948 Iter 19577, loss: 0.4423 Iter 19578, loss: 0.4706 Iter 19579, loss: 0.4790 Iter 19580, loss: 0.4825 Iter 19581, loss: 0.4456 Iter 19582, loss: 0.4757 Iter 19583, loss: 0.4800 Iter 19584, loss: 0.4805 Iter 19585, loss: 0.4689 Iter 19586, loss: 0.4503 Iter 19587, loss: 0.4830 Iter 19588, loss: 0.4835 Iter 19589, loss: 0.5004 Iter 19590, loss: 0.4369 Iter 19591, loss: 0.4643 Iter 19592, loss: 0.4445 Iter 19593, loss: 0.4563 Iter 19594, loss: 0.4999 Iter 19595, loss: 0.4725 Iter 19596, loss: 0.4864 Iter 19597, loss: 0.4579 Iter 19598, loss: 0.4686 Iter 19599, loss: 0.4692 Iter 19600, loss: 0.4693 Iter 19601, loss: 0.4544 Iter 19602, loss: 0.5069 Iter 19603, loss: 0.4404 Iter 19604, loss: 0.4895 Iter 19605, loss: 0.4475 Iter 19606, loss: 0.4973 Iter 19607, loss: 0.4915 Iter 19608, loss: 0.4307 Iter 19609, loss: 0.4635 Iter 19610, loss: 0.4651 Iter 19611, loss: 0.4868 Iter 19612, loss: 0.4552

Iter 19613, loss: 0.4920 Iter 19614, loss: 0.4195 Iter 19615, loss: 0.4855 Iter 19616, loss: 0.4867 Iter 19617, loss: 0.4677 Iter 19618, loss: 0.4641 Iter 19619, loss: 0.4206 Iter 19620, loss: 0.4616 Iter 19621, loss: 0.5276 Iter 19622, loss: 0.5138 Iter 19623, loss: 0.4781 Iter 19624, loss: 0.4825 Iter 19625, loss: 0.4696 Iter 19626, loss: 0.5083 Iter 19627, loss: 0.4855 Iter 19628, loss: 0.4418 Iter 19629, loss: 0.4595 Iter 19630, loss: 0.5167 Iter 19631, loss: 0.4905 Iter 19632, loss: 0.5282 Iter 19633, loss: 0.4920 Iter 19634, loss: 0.4846 Iter 19635, loss: 0.4996 Iter 19636, loss: 0.4707 Iter 19637, loss: 0.5012 Iter 19638, loss: 0.5060 Iter 19639, loss: 0.4824 Iter 19640, loss: 0.4372 Iter 19641, loss: 0.4685 Iter 19642, loss: 0.4703 Iter 19643, loss: 0.4874 Iter 19644, loss: 0.4950 Iter 19645, loss: 0.4909 Iter 19646, loss: 0.4511 Iter 19647, loss: 0.4957 Iter 19648, loss: 0.5027 Iter 19649, loss: 0.5028 Iter 19650, loss: 0.4849 Iter 19651, loss: 0.4724 Iter 19652, loss: 0.4351 Iter 19653, loss: 0.4930 Iter 19654, loss: 0.4267

Iter 19655, loss: 0.4979 Iter 19656, loss: 0.4877 Iter 19657, loss: 0.4740 Iter 19658, loss: 0.5087 Iter 19659, loss: 0.4771 Iter 19660, loss: 0.5202 Iter 19661, loss: 0.4321 Iter 19662, loss: 0.5273 Iter 19663, loss: 0.4877 Iter 19664, loss: 0.4588 Iter 19665, loss: 0.5125 Iter 19666, loss: 0.4776 Iter 19667, loss: 0.4839 Iter 19668, loss: 0.4911 Iter 19669, loss: 0.5024 Iter 19670, loss: 0.4744 Iter 19671, loss: 0.4553 Iter 19672, loss: 0.4414 Iter 19673, loss: 0.4993 Iter 19674, loss: 0.4502 Iter 19675, loss: 0.4626 Iter 19676, loss: 0.4524 Iter 19677, loss: 0.4654 Iter 19678, loss: 0.5249 Iter 19679, loss: 0.4819 Iter 19680, loss: 0.4911 Iter 19681, loss: 0.5026 Iter 19682, loss: 0.5079 Iter 19683, loss: 0.5049 Iter 19684, loss: 0.4487 Iter 19685, loss: 0.5038 Iter 19686, loss: 0.4555 Iter 19687, loss: 0.4469 Iter 19688, loss: 0.4877 Iter 19689, loss: 0.5151 Iter 19690, loss: 0.4655 Iter 19691, loss: 0.4781 Iter 19692, loss: 0.4704 Iter 19693, loss: 0.4677 Iter 19694, loss: 0.4649 Iter 19695, loss: 0.4507 Iter 19696, loss: 0.4749

Iter 19697, loss: 0.5009 Iter 19698, loss: 0.4404 Iter 19699, loss: 0.4618 Iter 19700, loss: 0.4800 Iter 19701, loss: 0.4254 Iter 19702, loss: 0.5037 Iter 19703, loss: 0.4871 Iter 19704, loss: 0.4821 Iter 19705, loss: 0.5056 Iter 19706, loss: 0.4550 Iter 19707, loss: 0.5109 Iter 19708, loss: 0.4772 Iter 19709, loss: 0.4766 Iter 19710, loss: 0.4747 Iter 19711, loss: 0.4723 Iter 19712, loss: 0.4968 Iter 19713, loss: 0.5099 Iter 19714, loss: 0.4776 Iter 19715, loss: 0.4610 Iter 19716, loss: 0.4229 Iter 19717, loss: 0.4399 Iter 19718, loss: 0.4909 Iter 19719, loss: 0.4652 Iter 19720, loss: 0.4719 Iter 19721, loss: 0.4723 Iter 19722, loss: 0.4797 Iter 19723, loss: 0.4702 Iter 19724, loss: 0.4576 Iter 19725, loss: 0.4733 Iter 19726, loss: 0.4752 Iter 19727, loss: 0.4688 Iter 19728, loss: 0.4766 Iter 19729, loss: 0.4449 Iter 19730, loss: 0.4628 Iter 19731, loss: 0.4773 Iter 19732, loss: 0.4601 Iter 19733, loss: 0.4769 Iter 19734, loss: 0.4964 Iter 19735, loss: 0.4971 Iter 19736, loss: 0.4698 Iter 19737, loss: 0.4397 Iter 19738, loss: 0.4911

Iter 19739, loss: 0.4929 Iter 19740, loss: 0.4325 Iter 19741, loss: 0.4926 Iter 19742, loss: 0.5224 Iter 19743, loss: 0.4970 Iter 19744, loss: 0.4805 Iter 19745, loss: 0.4937 Iter 19746, loss: 0.4777 Iter 19747, loss: 0.4563 Iter 19748, loss: 0.5037 Iter 19749, loss: 0.5048 Iter 19750, loss: 0.4617 Iter 19751, loss: 0.4866 Iter 19752, loss: 0.4706 Iter 19753, loss: 0.4634 Iter 19754, loss: 0.4628 Iter 19755, loss: 0.4568 Iter 19756, loss: 0.4827 Iter 19757, loss: 0.4442 Iter 19758, loss: 0.5195 Iter 19759, loss: 0.4918 Iter 19760, loss: 0.4520 Iter 19761, loss: 0.4445 Iter 19762, loss: 0.4929 Iter 19763, loss: 0.4453 Iter 19764, loss: 0.4557 Iter 19765, loss: 0.4766 Iter 19766, loss: 0.4635 Iter 19767, loss: 0.4777 Iter 19768, loss: 0.4518 Iter 19769, loss: 0.4863 Iter 19770, loss: 0.4783 Iter 19771, loss: 0.4580 Iter 19772, loss: 0.4636 Iter 19773, loss: 0.4460 Iter 19774, loss: 0.4640 Iter 19775, loss: 0.4597 Iter 19776, loss: 0.5105 Iter 19777, loss: 0.4589 Iter 19778, loss: 0.4990 Iter 19779, loss: 0.4709 Iter 19780, loss: 0.5024

Iter 19781, loss: 0.4832 Iter 19782, loss: 0.4997 Iter 19783, loss: 0.4662 Iter 19784, loss: 0.4623 Iter 19785, loss: 0.5124 Iter 19786, loss: 0.4566 Iter 19787, loss: 0.4993 Iter 19788, loss: 0.4366 Iter 19789, loss: 0.4773 Iter 19790, loss: 0.4702 Iter 19791, loss: 0.4588 Iter 19792, loss: 0.4581 Iter 19793, loss: 0.5089 Iter 19794, loss: 0.4390 Iter 19795, loss: 0.4912 Iter 19796, loss: 0.4744 Iter 19797, loss: 0.4538 Iter 19798, loss: 0.4506 Iter 19799, loss: 0.5012 Iter 19800, loss: 0.4903 Iter 19801, loss: 0.5062 Iter 19802, loss: 0.4583 Iter 19803, loss: 0.4662 Iter 19804, loss: 0.4937 Iter 19805, loss: 0.4736 Iter 19806, loss: 0.4566 Iter 19807, loss: 0.5129 Iter 19808, loss: 0.4437 Iter 19809, loss: 0.4685 Iter 19810, loss: 0.4843 Iter 19811, loss: 0.4302 Iter 19812, loss: 0.4800 Iter 19813, loss: 0.4831 Iter 19814, loss: 0.5016 Iter 19815, loss: 0.5040 Iter 19816, loss: 0.4772 Iter 19817, loss: 0.4338 Iter 19818, loss: 0.4911 Iter 19819, loss: 0.4665 Iter 19820, loss: 0.4789 Iter 19821, loss: 0.4467 Iter 19822, loss: 0.4602

Iter 19823, loss: 0.4666 Iter 19824, loss: 0.4516 Iter 19825, loss: 0.4662 Iter 19826, loss: 0.4530 Iter 19827, loss: 0.4459 Iter 19828, loss: 0.4971 Iter 19829, loss: 0.4350 Iter 19830, loss: 0.5416 Iter 19831, loss: 0.5149 Iter 19832, loss: 0.4751 Iter 19833, loss: 0.4903 Iter 19834, loss: 0.4594 Iter 19835, loss: 0.4922 Iter 19836, loss: 0.4687 Iter 19837, loss: 0.4705 Iter 19838, loss: 0.4684 Iter 19839, loss: 0.4690 Iter 19840, loss: 0.4825 Iter 19841, loss: 0.4651 Iter 19842, loss: 0.4980 Iter 19843, loss: 0.4623 Iter 19844, loss: 0.4822 Iter 19845, loss: 0.4747 Iter 19846, loss: 0.4678 Iter 19847, loss: 0.4630 Iter 19848, loss: 0.4637 Iter 19849, loss: 0.4645 Iter 19850, loss: 0.4999 Iter 19851, loss: 0.5188 Iter 19852, loss: 0.4949 Iter 19853, loss: 0.5017 Iter 19854, loss: 0.4993 Iter 19855, loss: 0.4654 Iter 19856, loss: 0.4699 Iter 19857, loss: 0.4566 Iter 19858, loss: 0.4501 Iter 19859, loss: 0.4762 Iter 19860, loss: 0.4917 Iter 19861, loss: 0.4819 Iter 19862, loss: 0.4926 Iter 19863, loss: 0.4886 Iter 19864, loss: 0.4789

Iter 19865, loss: 0.4808 Iter 19866, loss: 0.4667 Iter 19867, loss: 0.4518 Iter 19868, loss: 0.4867 Iter 19869, loss: 0.4793 Iter 19870, loss: 0.4913 Iter 19871, loss: 0.4822 Iter 19872, loss: 0.4944 Iter 19873, loss: 0.4776 Iter 19874, loss: 0.4713 Iter 19875, loss: 0.4948 Iter 19876, loss: 0.5040 Iter 19877, loss: 0.4624 Iter 19878, loss: 0.4978 Iter 19879, loss: 0.5077 Iter 19880, loss: 0.4484 Iter 19881, loss: 0.5501 Iter 19882, loss: 0.4657 Iter 19883, loss: 0.4936 Iter 19884, loss: 0.4736 Iter 19885, loss: 0.4490 Iter 19886, loss: 0.4511 Iter 19887, loss: 0.4767 Iter 19888, loss: 0.4864 Iter 19889, loss: 0.4761 Iter 19890, loss: 0.4956 Iter 19891, loss: 0.4904 Iter 19892, loss: 0.4471 Iter 19893, loss: 0.4532 Iter 19894, loss: 0.4608 Iter 19895, loss: 0.4477 Iter 19896, loss: 0.4965 Iter 19897, loss: 0.5198 Iter 19898, loss: 0.4559 Iter 19899, loss: 0.4590 Iter 19900, loss: 0.4691 Iter 19901, loss: 0.4582 Iter 19902, loss: 0.4695 Iter 19903, loss: 0.4741 Iter 19904, loss: 0.4280 Iter 19905, loss: 0.5057 Iter 19906, loss: 0.4864

Iter 19907, loss: 0.5135 Iter 19908, loss: 0.4685 Iter 19909, loss: 0.4617 Iter 19910, loss: 0.4790 Iter 19911, loss: 0.4564 Iter 19912, loss: 0.4401 Iter 19913, loss: 0.4994 Iter 19914, loss: 0.4713 Iter 19915, loss: 0.4962 Iter 19916, loss: 0.4759 Iter 19917, loss: 0.4815 Iter 19918, loss: 0.5098 Iter 19919, loss: 0.4892 Iter 19920, loss: 0.4533 Iter 19921, loss: 0.4727 Iter 19922, loss: 0.4903 Iter 19923, loss: 0.4757 Iter 19924, loss: 0.4310 Iter 19925, loss: 0.5056 Iter 19926, loss: 0.4969 Iter 19927, loss: 0.4506 Iter 19928, loss: 0.4918 Iter 19929, loss: 0.4864 Iter 19930, loss: 0.4848 Iter 19931, loss: 0.5013 Iter 19932, loss: 0.4868 Iter 19933, loss: 0.4646 Iter 19934, loss: 0.4733 Iter 19935, loss: 0.5103 Iter 19936, loss: 0.4592 Iter 19937, loss: 0.4762 Iter 19938, loss: 0.4828 Iter 19939, loss: 0.4631 Iter 19940, loss: 0.4665 Iter 19941, loss: 0.4759 Iter 19942, loss: 0.4362 Iter 19943, loss: 0.4978 Iter 19944, loss: 0.4597 Iter 19945, loss: 0.4925 Iter 19946, loss: 0.4515 Iter 19947, loss: 0.4651 Iter 19948, loss: 0.5057

Iter 19949, loss: 0.4653 Iter 19950, loss: 0.5016 Iter 19951, loss: 0.4699 Iter 19952, loss: 0.4866 Iter 19953, loss: 0.4627 Iter 19954, loss: 0.4748 Iter 19955, loss: 0.4539 Iter 19956, loss: 0.5110 Iter 19957, loss: 0.4797 Iter 19958, loss: 0.4615 Iter 19959, loss: 0.4588 Iter 19960, loss: 0.5021 Iter 19961, loss: 0.4962 Iter 19962, loss: 0.5081 Iter 19963, loss: 0.4634 Iter 19964, loss: 0.4548 Iter 19965, loss: 0.5109 Iter 19966, loss: 0.5018 Iter 19967, loss: 0.4772 Iter 19968, loss: 0.4561 Iter 19969, loss: 0.4629 Iter 19970, loss: 0.4929 Iter 19971, loss: 0.4469 Iter 19972, loss: 0.4496 Iter 19973, loss: 0.4658 Iter 19974, loss: 0.4364 Iter 19975, loss: 0.5124 Iter 19976, loss: 0.4837 Iter 19977, loss: 0.4931 Iter 19978, loss: 0.4657 Iter 19979, loss: 0.4530 Iter 19980, loss: 0.4746 Iter 19981, loss: 0.4800 Iter 19982, loss: 0.4604 Iter 19983, loss: 0.4751 Iter 19984, loss: 0.4682 Iter 19985, loss: 0.4467 Iter 19986, loss: 0.4532 Iter 19987, loss: 0.4682 Iter 19988, loss: 0.5075 Iter 19989, loss: 0.4648 Iter 19990, loss: 0.5184

Iter 19991, loss: 0.4696 Iter 19992, loss: 0.4734 Iter 19993, loss: 0.4730 Iter 19994, loss: 0.4521 Iter 19995, loss: 0.4488 Iter 19996, loss: 0.4802 Iter 19997, loss: 0.4899 Iter 19998, loss: 0.4607 Iter 19999, loss: 0.4738

Now we can visualize the diffused data along with the time dependent score model function:

fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(12, 12))

for idx, ax in enumerate(axes.flat):
    ax.scatter(x_diffused_example[:, idx, 0], x_diffused_example[:, idx, 1], s=1)
    ax.set_title(f"Step {idx}", fontdict={"size": 10}, pad=3)
    ax.set_xticks([])
    ax.set_yticks([])

    model_cond_at_time_t = partial(model_cond, t=jnp.array([idx]))
    plot_vector_field(model_cond_at_time_t, ax=ax, x_range=(-2, 2, 20), y_range=(-2, 2, 20), color="red")
    ax.set_xlim(-2, 2)
    ax.set_ylim(-2, 2)

plt.savefig("images/time-dependent-gradient-field.png", dpi=300);

png

You can see that the vector field now changes with time: at late steps it corresponds to a field with a single “sink”, where all sample are moved towards too, and than gradually there is more detailed structure introduced. In computational imaging this is also sometimes called a “coarse to fine” scheme.

Now let’s implement the sampling algorithm (aka “reverse process”). It is basically equivalent to the Langevin sampling, except that our “hyper-parameters” $\alpha_0$ and $p_0$ are replaced by well defined combinations of the $\beta$, $\alpha$ and $\bar{\alpha}$ parameters also describing the forward diffusion process. Also we have to make sure to actually reverse the process in time:

Args = namedtuple("Args", ["key", "idx", "x", "alphas", "betas", "alpha_bars"])



def sample(model, args, _):
    key, subkey = random.split(args.key)

    alpha, alpha_bar, beta = args.alphas[args.idx], args.alpha_bars[args.idx], args.betas[args.idx]

    model_cond_at_time_t = partial(model, t=jnp.array([args.idx]))
    dx = jax.vmap(model_cond_at_time_t)(x=args.x)
    
    eps = random.normal(subkey, args.x.shape)
    eps = jnp.where(args.idx == 0, 0, eps)
  
    x = 1. / jnp.sqrt(alpha) * (args.x - (beta / jnp.sqrt(1 - alpha_bar)) * dx) + jnp.sqrt(beta) * eps
    
    return Args(key, args.idx - 1, x, args.alphas, args.betas, args.alpha_bars), x


n_samples = 10_000

key = random.PRNGKey(42)
key, subkey = random.split(key)

init = Args(
    key=key,
    idx=n_t_steps - 1,
    x=random.normal(subkey, (n_samples, 2)),
    alphas=1 - betas,
    betas=betas,
    alpha_bars=jnp.cumprod(alphas),
)

result, sample_trace = jax.lax.scan(partial(sample, model_cond), init, length=n_t_steps)

Now let’s visualize the final samples:

def animate_trajectory(x, filename, x_range=(-2, 2), y_range=(-2, 2), **kwargs):
    """Animate diffusion trajectory"""
    fig, ax = plt.subplots()

    kwargs.setdefault("markersize", 0.2)
    kwargs.setdefault("linestyle", "")
    kwargs.setdefault("marker", "o")

    line, = ax.plot([], [], **kwargs)
    ax.set_aspect('equal')
    ax.set_xlim(*x_range)
    ax.set_ylim(*y_range)
    ax.set_xlabel('x')
    ax.set_ylabel('y')

    def animate(idx):
        line.set_data(x[idx, :, 0], x[idx, :, 1]) # Update line data
        return line,

    ani = animation.FuncAnimation(fig, animate, frames=len(x), interval=100, blit=True)
    ani.save(filename, writer='ffmpeg')

ax = plt.subplot()

idx = -1
ax.scatter(sample_trace[idx, :, 0], sample_trace[idx, :, 1], s=1)
ax.set_aspect('equal')
scale = 2
ax.set_xlim(-scale, scale)
ax.set_ylim(-scale, scale)
ax.set_xlabel("x")
ax.set_ylabel("y");

png

animate_trajectory(sample_trace, "images/ddpm-swiss-roll-animation.gif");

MovieWriter ffmpeg unavailable; using Pillow instead.

png

Here is what teh animation looks like:

animation

This looks much better than the earlier attempt without the time dependences!

Now we have covered all the necessary ideas for understanding diffusion models. The main advantage of diffusion models is the quality / accuray of the sampling. The main disadvantage is that the sampling is rather slow. To arrive at the final sample we evaluated the model as much times as we had time steps instead of as single evaluation. For “real” large scale model the nnumber of time steps can be as high as 1000 or more. However the working principle for the larger production models is exactly the same!

Here are some of the relevant differences:

Here are some links with larger scaled implementations

There is a nice connection between diffusion models and stochastic differential equations (SED): instead of the discreete time handling we introduced above one can handle the time as a continues variable and use OED solvers to handle the sampling process. A nice minimal examples is given in the Equinox documentation:

Thanks for the attention, I hope this was useful! If you have any further questions or comments please open an issue in this repo or reach out at one of the contacts given on my website.