challenging-america-word-gap-prediction-bigram/zad7.ipynb at master

s470618/challenging-america-word-gap-prediction-bigram

2023-06-30 14:08:09 +02:00

261 KiB

Raw Permalink Blame History

!pip install torchtext

Defaulting to user installation because normal site-packages is not writeable
Collecting torchtext
  Downloading torchtext-0.15.2-cp310-cp310-manylinux1_x86_64.whl (2.0 MB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m2.0/2.0 MB[0m [31m1.5 MB/s[0m eta [36m0:00:00[0m[36m0:00:01[0m[36m0:00:01[0m:01[0m
[?25hCollecting tqdm
  Using cached tqdm-4.65.0-py3-none-any.whl (77 kB)
Requirement already satisfied: numpy in /home/gedin/.local/lib/python3.10/site-packages (from torchtext) (1.24.3)
Collecting torchdata==0.6.1
  Downloading torchdata-0.6.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (4.6 MB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m4.6/4.6 MB[0m [31m1.5 MB/s[0m eta [36m0:00:00[0mm eta [36m0:00:01[0m[36m0:00:01[0m
[?25hRequirement already satisfied: requests in /usr/lib/python3/dist-packages (from torchtext) (2.25.1)
Collecting torch==2.0.1
  Downloading torch-2.0.1-cp310-cp310-manylinux1_x86_64.whl (619.9 MB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m619.9/619.9 MB[0m [31m1.1 MB/s[0m eta [36m0:00:00[0mm eta [36m0:00:01[0m[36m0:00:09[0m
[?25hCollecting sympy
  Downloading sympy-1.12-py3-none-any.whl (5.7 MB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m5.7/5.7 MB[0m [31m1.7 MB/s[0m eta [36m0:00:00[0mm eta [36m0:00:01[0m[36m0:00:01[0m
[?25hCollecting nvidia-cudnn-cu11==8.5.0.96
  Using cached nvidia_cudnn_cu11-8.5.0.96-2-py3-none-manylinux1_x86_64.whl (557.1 MB)
Collecting nvidia-cuda-cupti-cu11==11.7.101
  Using cached nvidia_cuda_cupti_cu11-11.7.101-py3-none-manylinux1_x86_64.whl (11.8 MB)
Collecting nvidia-cusparse-cu11==11.7.4.91
  Using cached nvidia_cusparse_cu11-11.7.4.91-py3-none-manylinux1_x86_64.whl (173.2 MB)
Collecting networkx
  Using cached networkx-3.1-py3-none-any.whl (2.1 MB)
Collecting nvidia-cufft-cu11==10.9.0.58
  Using cached nvidia_cufft_cu11-10.9.0.58-py3-none-manylinux1_x86_64.whl (168.4 MB)
Collecting filelock
  Downloading filelock-3.12.0-py3-none-any.whl (10 kB)
Collecting nvidia-cuda-runtime-cu11==11.7.99
  Using cached nvidia_cuda_runtime_cu11-11.7.99-py3-none-manylinux1_x86_64.whl (849 kB)
Collecting triton==2.0.0
  Downloading triton-2.0.0-1-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (63.3 MB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m63.3/63.3 MB[0m [31m1.3 MB/s[0m eta [36m0:00:00[0mm eta [36m0:00:01[0m[36m0:00:02[0m
[?25hCollecting nvidia-cusolver-cu11==11.4.0.1
  Using cached nvidia_cusolver_cu11-11.4.0.1-2-py3-none-manylinux1_x86_64.whl (102.6 MB)
Requirement already satisfied: jinja2 in /home/gedin/.local/lib/python3.10/site-packages (from torch==2.0.1->torchtext) (3.1.2)
Collecting nvidia-cublas-cu11==11.10.3.66
  Using cached nvidia_cublas_cu11-11.10.3.66-py3-none-manylinux1_x86_64.whl (317.1 MB)
Collecting typing-extensions
  Downloading typing_extensions-4.6.3-py3-none-any.whl (31 kB)
Collecting nvidia-nccl-cu11==2.14.3
  Using cached nvidia_nccl_cu11-2.14.3-py3-none-manylinux1_x86_64.whl (177.1 MB)
Collecting nvidia-cuda-nvrtc-cu11==11.7.99
  Using cached nvidia_cuda_nvrtc_cu11-11.7.99-2-py3-none-manylinux1_x86_64.whl (21.0 MB)
Collecting nvidia-curand-cu11==10.2.10.91
  Using cached nvidia_curand_cu11-10.2.10.91-py3-none-manylinux1_x86_64.whl (54.6 MB)
Collecting nvidia-nvtx-cu11==11.7.91
  Using cached nvidia_nvtx_cu11-11.7.91-py3-none-manylinux1_x86_64.whl (98 kB)
Requirement already satisfied: urllib3>=1.25 in /usr/lib/python3/dist-packages (from torchdata==0.6.1->torchtext) (1.26.5)
Requirement already satisfied: wheel in /usr/lib/python3/dist-packages (from nvidia-cublas-cu11==11.10.3.66->torch==2.0.1->torchtext) (0.37.1)
Requirement already satisfied: setuptools in /usr/lib/python3/dist-packages (from nvidia-cublas-cu11==11.10.3.66->torch==2.0.1->torchtext) (59.6.0)
Collecting lit
  Downloading lit-16.0.5.tar.gz (138 kB)
[2K     [38;2;114;156;31m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[0m [32m138.0/138.0 KB[0m [31m1.7 MB/s[0m eta [36m0:00:00[0m[31m1.6 MB/s[0m eta [36m0:00:01[0m
[?25h  Preparing metadata (setup.py) ... [?25ldone
[?25hCollecting cmake
  Using cached cmake-3.26.3-py2.py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl (24.0 MB)
Requirement already satisfied: MarkupSafe>=2.0 in /usr/lib/python3/dist-packages (from jinja2->torch==2.0.1->torchtext) (2.0.1)
Collecting mpmath>=0.19
  Using cached mpmath-1.3.0-py3-none-any.whl (536 kB)
Building wheels for collected packages: lit
  Building wheel for lit (setup.py) ... [?25ldone
[?25h  Created wheel for lit: filename=lit-16.0.5-py3-none-any.whl size=88192 sha256=f6c57a31a147cbfe0af3d6bf4b856390ad14c28a9ddb38c8044ec29331b35c26
  Stored in directory: /home/gedin/.cache/pip/wheels/eb/02/84/d82f0b1a6098209edf7e3607be6cc592ebbc015a8a3127c68d
Successfully built lit
Installing collected packages: mpmath, lit, cmake, typing-extensions, tqdm, sympy, nvidia-nvtx-cu11, nvidia-nccl-cu11, nvidia-cusparse-cu11, nvidia-curand-cu11, nvidia-cufft-cu11, nvidia-cuda-runtime-cu11, nvidia-cuda-nvrtc-cu11, nvidia-cuda-cupti-cu11, nvidia-cublas-cu11, networkx, filelock, nvidia-cusolver-cu11, nvidia-cudnn-cu11, triton, torch, torchdata, torchtext
Successfully installed cmake-3.26.3 filelock-3.12.0 lit-16.0.5 mpmath-1.3.0 networkx-3.1 nvidia-cublas-cu11-11.10.3.66 nvidia-cuda-cupti-cu11-11.7.101 nvidia-cuda-nvrtc-cu11-11.7.99 nvidia-cuda-runtime-cu11-11.7.99 nvidia-cudnn-cu11-8.5.0.96 nvidia-cufft-cu11-10.9.0.58 nvidia-curand-cu11-10.2.10.91 nvidia-cusolver-cu11-11.4.0.1 nvidia-cusparse-cu11-11.7.4.91 nvidia-nccl-cu11-2.14.3 nvidia-nvtx-cu11-11.7.91 sympy-1.12 torch-2.0.1 torchdata-0.6.1 torchtext-0.15.2 tqdm-4.65.0 triton-2.0.0 typing-extensions-4.6.3

train_file ='train/in.tsv.xz'
test_file = 'dev-0/in.tsv.xz'
out_file = 'dev-0/out.tsv'

from itertools import islice
import regex as re
import sys
from torchtext.vocab import build_vocab_from_iterator
import lzma
import pickle
import re
import torch
from torch import nn
from torch.utils.data import IterableDataset
import itertools
from torch.utils.data import DataLoader
import gc

embed_size = 300
device = 'cuda'
vocab_size = 25000
batch_s = 3200
learning_rate = 0.0001
epochs = 4
k = 20 #top k words
wildcard_minweight = 0.1

###preprocessing
def preprocess(line):
    line = get_rid_of_header(line)
    line = replace_endline(line)
    return line

def get_rid_of_header(line):
    line = line.split('\t')[6:]
    return "".join(line)
    
def replace_endline(line):
    line = line.replace("\\\\n", " ")
    return line


def get_last_word(text):
    """Return the last word of a string."""
    last_word = ""
    for i in range(len(text)-1, -1, -1):
        if text[i] == ' ':
            return last_word[::-1].rstrip()
        else:
            last_word += text[i]
    return last_word[::-1].rstrip()

def get_first_word(text):
    """Return the first word of a string."""
    word = ""
    for i in range(len(text)-1):
        if text[i] == ' ':
            return word
        else:
            word += text[i]
    return word


def get_words_from_line(line):
    line = line.rstrip()
    yield '<s>'
    line = preprocess(line)
    for t in line.split(' '):
        yield t
    yield '</s>'


def get_word_lines_from_file(file_name):
    n = 0
    with lzma.open(file_name, 'r') as fh:
        for line in fh:
            n+=1
            if n%1000==0:
                print(n)
            yield get_words_from_line(line.decode('utf-8'))

vocab = build_vocab_from_iterator(
    get_word_lines_from_file(train_file),
    max_tokens = vocab_size,
    specials = ['<unk>'])

with open('filename.pickle', 'wb') as handle:
    pickle.dump(vocab, handle, protocol=pickle.HIGHEST_PROTOCOL)

vocab.lookup_tokens([0, 1, 2, 10, 2000])

['<unk>', 'the', 'of', 'was', 'ladies']

Definicja sieci

Naszą prostą sieć neuronową zaimplementujemy używając frameworku PyTorch.

class SimpleBigramNeuralLanguageModel(nn.Module):
  def __init__(self, vocabulary_size, embedding_size):
      super(SimpleBigramNeuralLanguageModel, self).__init__()
      self.model = nn.Sequential(
          nn.Embedding(vocabulary_size, embedding_size),
          nn.Linear(embedding_size, vocabulary_size),
          nn.Softmax()
      )
    
  def forward(self, x):
      return self.model(x)

with open('filename.pickle','rb') as handle:
    vocab = pickle.load(handle)

vocab.set_default_index(vocab['<unk>'])

help(vocab)

Help on Vocab in module torchtext.vocab.vocab object:

class Vocab(torch.nn.modules.module.Module)
 |  Vocab(vocab) -> None
 |  
 |  Base class for all neural network modules.
 |  
 |  Your models should also subclass this class.
 |  
 |  Modules can also contain other Modules, allowing to nest them in
 |  a tree structure. You can assign the submodules as regular attributes::
 |  
 |      import torch.nn as nn
 |      import torch.nn.functional as F
 |  
 |      class Model(nn.Module):
 |          def __init__(self):
 |              super().__init__()
 |              self.conv1 = nn.Conv2d(1, 20, 5)
 |              self.conv2 = nn.Conv2d(20, 20, 5)
 |  
 |          def forward(self, x):
 |              x = F.relu(self.conv1(x))
 |              return F.relu(self.conv2(x))
 |  
 |  Submodules assigned in this way will be registered, and will have their
 |  parameters converted too when you call :meth:`to`, etc.
 |  
 |  .. note::
 |      As per the example above, an ``__init__()`` call to the parent class
 |      must be made before assignment on the child.
 |  
 |  :ivar training: Boolean represents whether this module is in training or
 |                  evaluation mode.
 |  :vartype training: bool
 |  
 |  Method resolution order:
 |      Vocab
 |      torch.nn.modules.module.Module
 |      builtins.object
 |  
 |  Methods defined here:
 |  
 |  __contains__(self, token: str) -> bool
 |      Args:
 |          token: The token for which to check the membership.
 |      
 |      Returns:
 |          Whether the token is member of vocab or not.
 |  
 |  __getitem__(self, token: str) -> int
 |      Args:
 |          token: The token used to lookup the corresponding index.
 |      
 |      Returns:
 |          The index corresponding to the associated token.
 |  
 |  __init__(self, vocab) -> None
 |      Initializes internal Module state, shared by both nn.Module and ScriptModule.
 |  
 |  __len__(self) -> int
 |      Returns:
 |          The length of the vocab.
 |  
 |  __prepare_scriptable__(self)
 |      Return a JITable Vocab.
 |  
 |  append_token(self, token: str) -> None
 |      Args:
 |          token: The token used to lookup the corresponding index.
 |      
 |      Raises:
 |          RuntimeError: If `token` already exists in the vocab
 |  
 |  forward(self, tokens: List[str]) -> List[int]
 |      Calls the `lookup_indices` method
 |      
 |      Args:
 |          tokens: a list of tokens used to lookup their corresponding `indices`.
 |      
 |      Returns:
 |          The indices associated with a list of `tokens`.
 |  
 |  get_default_index(self) -> Union[int, NoneType]
 |      Returns:
 |          Value of default index if it is set.
 |  
 |  get_itos(self) -> List[str]
 |      Returns:
 |          List mapping indices to tokens.
 |  
 |  get_stoi(self) -> Dict[str, int]
 |      Returns:
 |          Dictionary mapping tokens to indices.
 |  
 |  insert_token(self, token: str, index: int) -> None
 |      Args:
 |          token: The token used to lookup the corresponding index.
 |          index: The index corresponding to the associated token.
 |      Raises:
 |          RuntimeError: If `index` is not in range [0, Vocab.size()] or if `token` already exists in the vocab.
 |  
 |  lookup_indices(self, tokens: List[str]) -> List[int]
 |      Args:
 |          tokens: the tokens used to lookup their corresponding `indices`.
 |      
 |      Returns:
 |          The 'indices` associated with `tokens`.
 |  
 |  lookup_token(self, index: int) -> str
 |      Args:
 |          index: The index corresponding to the associated token.
 |      
 |      Returns:
 |          token: The token used to lookup the corresponding index.
 |      
 |      Raises:
 |          RuntimeError: If `index` not in range [0, itos.size()).
 |  
 |  lookup_tokens(self, indices: List[int]) -> List[str]
 |      Args:
 |          indices: The `indices` used to lookup their corresponding`tokens`.
 |      
 |      Returns:
 |          The `tokens` associated with `indices`.
 |      
 |      Raises:
 |          RuntimeError: If an index within `indices` is not int range [0, itos.size()).
 |  
 |  set_default_index(self, index: Union[int, NoneType]) -> None
 |      Args:
 |          index: Value of default index. This index will be returned when OOV token is queried.
 |  
 |  ----------------------------------------------------------------------
 |  Readonly properties defined here:
 |  
 |  is_jitable
 |  
 |  ----------------------------------------------------------------------
 |  Data and other attributes defined here:
 |  
 |  __jit_unused_properties__ = ['is_jitable']
 |  
 |  ----------------------------------------------------------------------
 |  Methods inherited from torch.nn.modules.module.Module:
 |  
 |  __call__ = _call_impl(self, *args, **kwargs)
 |  
 |  __delattr__(self, name)
 |      Implement delattr(self, name).
 |  
 |  __dir__(self)
 |      Default dir() implementation.
 |  
 |  __getattr__(self, name: str) -> Union[torch.Tensor, ForwardRef('Module')]
 |  
 |  __repr__(self)
 |      Return repr(self).
 |  
 |  __setattr__(self, name: str, value: Union[torch.Tensor, ForwardRef('Module')]) -> None
 |      Implement setattr(self, name, value).
 |  
 |  __setstate__(self, state)
 |  
 |  add_module(self, name: str, module: Union[ForwardRef('Module'), NoneType]) -> None
 |      Adds a child module to the current module.
 |      
 |      The module can be accessed as an attribute using the given name.
 |      
 |      Args:
 |          name (str): name of the child module. The child module can be
 |              accessed from this module using the given name
 |          module (Module): child module to be added to the module.
 |  
 |  apply(self: ~T, fn: Callable[[ForwardRef('Module')], NoneType]) -> ~T
 |      Applies ``fn`` recursively to every submodule (as returned by ``.children()``)
 |      as well as self. Typical use includes initializing the parameters of a model
 |      (see also :ref:`nn-init-doc`).
 |      
 |      Args:
 |          fn (:class:`Module` -> None): function to be applied to each submodule
 |      
 |      Returns:
 |          Module: self
 |      
 |      Example::
 |      
 |          >>> @torch.no_grad()
 |          >>> def init_weights(m):
 |          >>>     print(m)
 |          >>>     if type(m) == nn.Linear:
 |          >>>         m.weight.fill_(1.0)
 |          >>>         print(m.weight)
 |          >>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))
 |          >>> net.apply(init_weights)
 |          Linear(in_features=2, out_features=2, bias=True)
 |          Parameter containing:
 |          tensor([[1., 1.],
 |                  [1., 1.]], requires_grad=True)
 |          Linear(in_features=2, out_features=2, bias=True)
 |          Parameter containing:
 |          tensor([[1., 1.],
 |                  [1., 1.]], requires_grad=True)
 |          Sequential(
 |            (0): Linear(in_features=2, out_features=2, bias=True)
 |            (1): Linear(in_features=2, out_features=2, bias=True)
 |          )
 |  
 |  bfloat16(self: ~T) -> ~T
 |      Casts all floating point parameters and buffers to ``bfloat16`` datatype.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Returns:
 |          Module: self
 |  
 |  buffers(self, recurse: bool = True) -> Iterator[torch.Tensor]
 |      Returns an iterator over module buffers.
 |      
 |      Args:
 |          recurse (bool): if True, then yields buffers of this module
 |              and all submodules. Otherwise, yields only buffers that
 |              are direct members of this module.
 |      
 |      Yields:
 |          torch.Tensor: module buffer
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> for buf in model.buffers():
 |          >>>     print(type(buf), buf.size())
 |          <class 'torch.Tensor'> (20L,)
 |          <class 'torch.Tensor'> (20L, 1L, 5L, 5L)
 |  
 |  children(self) -> Iterator[ForwardRef('Module')]
 |      Returns an iterator over immediate children modules.
 |      
 |      Yields:
 |          Module: a child module
 |  
 |  cpu(self: ~T) -> ~T
 |      Moves all model parameters and buffers to the CPU.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Returns:
 |          Module: self
 |  
 |  cuda(self: ~T, device: Union[int, torch.device, NoneType] = None) -> ~T
 |      Moves all model parameters and buffers to the GPU.
 |      
 |      This also makes associated parameters and buffers different objects. So
 |      it should be called before constructing optimizer if the module will
 |      live on GPU while being optimized.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Args:
 |          device (int, optional): if specified, all parameters will be
 |              copied to that device
 |      
 |      Returns:
 |          Module: self
 |  
 |  double(self: ~T) -> ~T
 |      Casts all floating point parameters and buffers to ``double`` datatype.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Returns:
 |          Module: self
 |  
 |  eval(self: ~T) -> ~T
 |      Sets the module in evaluation mode.
 |      
 |      This has any effect only on certain modules. See documentations of
 |      particular modules for details of their behaviors in training/evaluation
 |      mode, if they are affected, e.g. :class:`Dropout`, :class:`BatchNorm`,
 |      etc.
 |      
 |      This is equivalent with :meth:`self.train(False) <torch.nn.Module.train>`.
 |      
 |      See :ref:`locally-disable-grad-doc` for a comparison between
 |      `.eval()` and several similar mechanisms that may be confused with it.
 |      
 |      Returns:
 |          Module: self
 |  
 |  extra_repr(self) -> str
 |      Set the extra representation of the module
 |      
 |      To print customized extra information, you should re-implement
 |      this method in your own modules. Both single-line and multi-line
 |      strings are acceptable.
 |  
 |  float(self: ~T) -> ~T
 |      Casts all floating point parameters and buffers to ``float`` datatype.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Returns:
 |          Module: self
 |  
 |  get_buffer(self, target: str) -> 'Tensor'
 |      Returns the buffer given by ``target`` if it exists,
 |      otherwise throws an error.
 |      
 |      See the docstring for ``get_submodule`` for a more detailed
 |      explanation of this method's functionality as well as how to
 |      correctly specify ``target``.
 |      
 |      Args:
 |          target: The fully-qualified string name of the buffer
 |              to look for. (See ``get_submodule`` for how to specify a
 |              fully-qualified string.)
 |      
 |      Returns:
 |          torch.Tensor: The buffer referenced by ``target``
 |      
 |      Raises:
 |          AttributeError: If the target string references an invalid
 |              path or resolves to something that is not a
 |              buffer
 |  
 |  get_extra_state(self) -> Any
 |      Returns any extra state to include in the module's state_dict.
 |      Implement this and a corresponding :func:`set_extra_state` for your module
 |      if you need to store extra state. This function is called when building the
 |      module's `state_dict()`.
 |      
 |      Note that extra state should be picklable to ensure working serialization
 |      of the state_dict. We only provide provide backwards compatibility guarantees
 |      for serializing Tensors; other objects may break backwards compatibility if
 |      their serialized pickled form changes.
 |      
 |      Returns:
 |          object: Any extra state to store in the module's state_dict
 |  
 |  get_parameter(self, target: str) -> 'Parameter'
 |      Returns the parameter given by ``target`` if it exists,
 |      otherwise throws an error.
 |      
 |      See the docstring for ``get_submodule`` for a more detailed
 |      explanation of this method's functionality as well as how to
 |      correctly specify ``target``.
 |      
 |      Args:
 |          target: The fully-qualified string name of the Parameter
 |              to look for. (See ``get_submodule`` for how to specify a
 |              fully-qualified string.)
 |      
 |      Returns:
 |          torch.nn.Parameter: The Parameter referenced by ``target``
 |      
 |      Raises:
 |          AttributeError: If the target string references an invalid
 |              path or resolves to something that is not an
 |              ``nn.Parameter``
 |  
 |  get_submodule(self, target: str) -> 'Module'
 |      Returns the submodule given by ``target`` if it exists,
 |      otherwise throws an error.
 |      
 |      For example, let's say you have an ``nn.Module`` ``A`` that
 |      looks like this:
 |      
 |      .. code-block:: text
 |      
 |          A(
 |              (net_b): Module(
 |                  (net_c): Module(
 |                      (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))
 |                  )
 |                  (linear): Linear(in_features=100, out_features=200, bias=True)
 |              )
 |          )
 |      
 |      (The diagram shows an ``nn.Module`` ``A``. ``A`` has a nested
 |      submodule ``net_b``, which itself has two submodules ``net_c``
 |      and ``linear``. ``net_c`` then has a submodule ``conv``.)
 |      
 |      To check whether or not we have the ``linear`` submodule, we
 |      would call ``get_submodule("net_b.linear")``. To check whether
 |      we have the ``conv`` submodule, we would call
 |      ``get_submodule("net_b.net_c.conv")``.
 |      
 |      The runtime of ``get_submodule`` is bounded by the degree
 |      of module nesting in ``target``. A query against
 |      ``named_modules`` achieves the same result, but it is O(N) in
 |      the number of transitive modules. So, for a simple check to see
 |      if some submodule exists, ``get_submodule`` should always be
 |      used.
 |      
 |      Args:
 |          target: The fully-qualified string name of the submodule
 |              to look for. (See above example for how to specify a
 |              fully-qualified string.)
 |      
 |      Returns:
 |          torch.nn.Module: The submodule referenced by ``target``
 |      
 |      Raises:
 |          AttributeError: If the target string references an invalid
 |              path or resolves to something that is not an
 |              ``nn.Module``
 |  
 |  half(self: ~T) -> ~T
 |      Casts all floating point parameters and buffers to ``half`` datatype.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Returns:
 |          Module: self
 |  
 |  ipu(self: ~T, device: Union[int, torch.device, NoneType] = None) -> ~T
 |      Moves all model parameters and buffers to the IPU.
 |      
 |      This also makes associated parameters and buffers different objects. So
 |      it should be called before constructing optimizer if the module will
 |      live on IPU while being optimized.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Arguments:
 |          device (int, optional): if specified, all parameters will be
 |              copied to that device
 |      
 |      Returns:
 |          Module: self
 |  
 |  load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True)
 |      Copies parameters and buffers from :attr:`state_dict` into
 |      this module and its descendants. If :attr:`strict` is ``True``, then
 |      the keys of :attr:`state_dict` must exactly match the keys returned
 |      by this module's :meth:`~torch.nn.Module.state_dict` function.
 |      
 |      Args:
 |          state_dict (dict): a dict containing parameters and
 |              persistent buffers.
 |          strict (bool, optional): whether to strictly enforce that the keys
 |              in :attr:`state_dict` match the keys returned by this module's
 |              :meth:`~torch.nn.Module.state_dict` function. Default: ``True``
 |      
 |      Returns:
 |          ``NamedTuple`` with ``missing_keys`` and ``unexpected_keys`` fields:
 |              * **missing_keys** is a list of str containing the missing keys
 |              * **unexpected_keys** is a list of str containing the unexpected keys
 |      
 |      Note:
 |          If a parameter or buffer is registered as ``None`` and its corresponding key
 |          exists in :attr:`state_dict`, :meth:`load_state_dict` will raise a
 |          ``RuntimeError``.
 |  
 |  modules(self) -> Iterator[ForwardRef('Module')]
 |      Returns an iterator over all modules in the network.
 |      
 |      Yields:
 |          Module: a module in the network
 |      
 |      Note:
 |          Duplicate modules are returned only once. In the following
 |          example, ``l`` will be returned only once.
 |      
 |      Example::
 |      
 |          >>> l = nn.Linear(2, 2)
 |          >>> net = nn.Sequential(l, l)
 |          >>> for idx, m in enumerate(net.modules()):
 |          ...     print(idx, '->', m)
 |      
 |          0 -> Sequential(
 |            (0): Linear(in_features=2, out_features=2, bias=True)
 |            (1): Linear(in_features=2, out_features=2, bias=True)
 |          )
 |          1 -> Linear(in_features=2, out_features=2, bias=True)
 |  
 |  named_buffers(self, prefix: str = '', recurse: bool = True, remove_duplicate: bool = True) -> Iterator[Tuple[str, torch.Tensor]]
 |      Returns an iterator over module buffers, yielding both the
 |      name of the buffer as well as the buffer itself.
 |      
 |      Args:
 |          prefix (str): prefix to prepend to all buffer names.
 |          recurse (bool, optional): if True, then yields buffers of this module
 |              and all submodules. Otherwise, yields only buffers that
 |              are direct members of this module. Defaults to True.
 |          remove_duplicate (bool, optional): whether to remove the duplicated buffers in the result. Defaults to True.
 |      
 |      Yields:
 |          (str, torch.Tensor): Tuple containing the name and buffer
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> for name, buf in self.named_buffers():
 |          >>>     if name in ['running_var']:
 |          >>>         print(buf.size())
 |  
 |  named_children(self) -> Iterator[Tuple[str, ForwardRef('Module')]]
 |      Returns an iterator over immediate children modules, yielding both
 |      the name of the module as well as the module itself.
 |      
 |      Yields:
 |          (str, Module): Tuple containing a name and child module
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> for name, module in model.named_children():
 |          >>>     if name in ['conv4', 'conv5']:
 |          >>>         print(module)
 |  
 |  named_modules(self, memo: Union[Set[ForwardRef('Module')], NoneType] = None, prefix: str = '', remove_duplicate: bool = True)
 |      Returns an iterator over all modules in the network, yielding
 |      both the name of the module as well as the module itself.
 |      
 |      Args:
 |          memo: a memo to store the set of modules already added to the result
 |          prefix: a prefix that will be added to the name of the module
 |          remove_duplicate: whether to remove the duplicated module instances in the result
 |              or not
 |      
 |      Yields:
 |          (str, Module): Tuple of name and module
 |      
 |      Note:
 |          Duplicate modules are returned only once. In the following
 |          example, ``l`` will be returned only once.
 |      
 |      Example::
 |      
 |          >>> l = nn.Linear(2, 2)
 |          >>> net = nn.Sequential(l, l)
 |          >>> for idx, m in enumerate(net.named_modules()):
 |          ...     print(idx, '->', m)
 |      
 |          0 -> ('', Sequential(
 |            (0): Linear(in_features=2, out_features=2, bias=True)
 |            (1): Linear(in_features=2, out_features=2, bias=True)
 |          ))
 |          1 -> ('0', Linear(in_features=2, out_features=2, bias=True))
 |  
 |  named_parameters(self, prefix: str = '', recurse: bool = True, remove_duplicate: bool = True) -> Iterator[Tuple[str, torch.nn.parameter.Parameter]]
 |      Returns an iterator over module parameters, yielding both the
 |      name of the parameter as well as the parameter itself.
 |      
 |      Args:
 |          prefix (str): prefix to prepend to all parameter names.
 |          recurse (bool): if True, then yields parameters of this module
 |              and all submodules. Otherwise, yields only parameters that
 |              are direct members of this module.
 |          remove_duplicate (bool, optional): whether to remove the duplicated
 |              parameters in the result. Defaults to True.
 |      
 |      Yields:
 |          (str, Parameter): Tuple containing the name and parameter
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> for name, param in self.named_parameters():
 |          >>>     if name in ['bias']:
 |          >>>         print(param.size())
 |  
 |  parameters(self, recurse: bool = True) -> Iterator[torch.nn.parameter.Parameter]
 |      Returns an iterator over module parameters.
 |      
 |      This is typically passed to an optimizer.
 |      
 |      Args:
 |          recurse (bool): if True, then yields parameters of this module
 |              and all submodules. Otherwise, yields only parameters that
 |              are direct members of this module.
 |      
 |      Yields:
 |          Parameter: module parameter
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> for param in model.parameters():
 |          >>>     print(type(param), param.size())
 |          <class 'torch.Tensor'> (20L,)
 |          <class 'torch.Tensor'> (20L, 1L, 5L, 5L)
 |  
 |  register_backward_hook(self, hook: Callable[[ForwardRef('Module'), Union[Tuple[torch.Tensor, ...], torch.Tensor], Union[Tuple[torch.Tensor, ...], torch.Tensor]], Union[NoneType, Tuple[torch.Tensor, ...], torch.Tensor]]) -> torch.utils.hooks.RemovableHandle
 |      Registers a backward hook on the module.
 |      
 |      This function is deprecated in favor of :meth:`~torch.nn.Module.register_full_backward_hook` and
 |      the behavior of this function will change in future versions.
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_buffer(self, name: str, tensor: Union[torch.Tensor, NoneType], persistent: bool = True) -> None
 |      Adds a buffer to the module.
 |      
 |      This is typically used to register a buffer that should not to be
 |      considered a model parameter. For example, BatchNorm's ``running_mean``
 |      is not a parameter, but is part of the module's state. Buffers, by
 |      default, are persistent and will be saved alongside parameters. This
 |      behavior can be changed by setting :attr:`persistent` to ``False``. The
 |      only difference between a persistent buffer and a non-persistent buffer
 |      is that the latter will not be a part of this module's
 |      :attr:`state_dict`.
 |      
 |      Buffers can be accessed as attributes using given names.
 |      
 |      Args:
 |          name (str): name of the buffer. The buffer can be accessed
 |              from this module using the given name
 |          tensor (Tensor or None): buffer to be registered. If ``None``, then operations
 |              that run on buffers, such as :attr:`cuda`, are ignored. If ``None``,
 |              the buffer is **not** included in the module's :attr:`state_dict`.
 |          persistent (bool): whether the buffer is part of this module's
 |              :attr:`state_dict`.
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> self.register_buffer('running_mean', torch.zeros(num_features))
 |  
 |  register_forward_hook(self, hook: Union[Callable[[~T, Tuple[Any, ...], Any], Union[Any, NoneType]], Callable[[~T, Tuple[Any, ...], Dict[str, Any], Any], Union[Any, NoneType]]], *, prepend: bool = False, with_kwargs: bool = False) -> torch.utils.hooks.RemovableHandle
 |      Registers a forward hook on the module.
 |      
 |      The hook will be called every time after :func:`forward` has computed an output.
 |      
 |      If ``with_kwargs`` is ``False`` or not specified, the input contains only
 |      the positional arguments given to the module. Keyword arguments won't be
 |      passed to the hooks and only to the ``forward``. The hook can modify the
 |      output. It can modify the input inplace but it will not have effect on
 |      forward since this is called after :func:`forward` is called. The hook
 |      should have the following signature::
 |      
 |          hook(module, args, output) -> None or modified output
 |      
 |      If ``with_kwargs`` is ``True``, the forward hook will be passed the
 |      ``kwargs`` given to the forward function and be expected to return the
 |      output possibly modified. The hook should have the following signature::
 |      
 |          hook(module, args, kwargs, output) -> None or modified output
 |      
 |      Args:
 |          hook (Callable): The user defined hook to be registered.
 |          prepend (bool): If ``True``, the provided ``hook`` will be fired
 |              before all existing ``forward`` hooks on this
 |              :class:`torch.nn.modules.Module`. Otherwise, the provided
 |              ``hook`` will be fired after all existing ``forward`` hooks on
 |              this :class:`torch.nn.modules.Module`. Note that global
 |              ``forward`` hooks registered with
 |              :func:`register_module_forward_hook` will fire before all hooks
 |              registered by this method.
 |              Default: ``False``
 |          with_kwargs (bool): If ``True``, the ``hook`` will be passed the
 |              kwargs given to the forward function.
 |              Default: ``False``
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_forward_pre_hook(self, hook: Union[Callable[[~T, Tuple[Any, ...]], Union[Any, NoneType]], Callable[[~T, Tuple[Any, ...], Dict[str, Any]], Union[Tuple[Any, Dict[str, Any]], NoneType]]], *, prepend: bool = False, with_kwargs: bool = False) -> torch.utils.hooks.RemovableHandle
 |      Registers a forward pre-hook on the module.
 |      
 |      The hook will be called every time before :func:`forward` is invoked.
 |      
 |      
 |      If ``with_kwargs`` is false or not specified, the input contains only
 |      the positional arguments given to the module. Keyword arguments won't be
 |      passed to the hooks and only to the ``forward``. The hook can modify the
 |      input. User can either return a tuple or a single modified value in the
 |      hook. We will wrap the value into a tuple if a single value is returned
 |      (unless that value is already a tuple). The hook should have the
 |      following signature::
 |      
 |          hook(module, args) -> None or modified input
 |      
 |      If ``with_kwargs`` is true, the forward pre-hook will be passed the
 |      kwargs given to the forward function. And if the hook modifies the
 |      input, both the args and kwargs should be returned. The hook should have
 |      the following signature::
 |      
 |          hook(module, args, kwargs) -> None or a tuple of modified input and kwargs
 |      
 |      Args:
 |          hook (Callable): The user defined hook to be registered.
 |          prepend (bool): If true, the provided ``hook`` will be fired before
 |              all existing ``forward_pre`` hooks on this
 |              :class:`torch.nn.modules.Module`. Otherwise, the provided
 |              ``hook`` will be fired after all existing ``forward_pre`` hooks
 |              on this :class:`torch.nn.modules.Module`. Note that global
 |              ``forward_pre`` hooks registered with
 |              :func:`register_module_forward_pre_hook` will fire before all
 |              hooks registered by this method.
 |              Default: ``False``
 |          with_kwargs (bool): If true, the ``hook`` will be passed the kwargs
 |              given to the forward function.
 |              Default: ``False``
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_full_backward_hook(self, hook: Callable[[ForwardRef('Module'), Union[Tuple[torch.Tensor, ...], torch.Tensor], Union[Tuple[torch.Tensor, ...], torch.Tensor]], Union[NoneType, Tuple[torch.Tensor, ...], torch.Tensor]], prepend: bool = False) -> torch.utils.hooks.RemovableHandle
 |      Registers a backward hook on the module.
 |      
 |      The hook will be called every time the gradients with respect to a module
 |      are computed, i.e. the hook will execute if and only if the gradients with
 |      respect to module outputs are computed. The hook should have the following
 |      signature::
 |      
 |          hook(module, grad_input, grad_output) -> tuple(Tensor) or None
 |      
 |      The :attr:`grad_input` and :attr:`grad_output` are tuples that contain the gradients
 |      with respect to the inputs and outputs respectively. The hook should
 |      not modify its arguments, but it can optionally return a new gradient with
 |      respect to the input that will be used in place of :attr:`grad_input` in
 |      subsequent computations. :attr:`grad_input` will only correspond to the inputs given
 |      as positional arguments and all kwarg arguments are ignored. Entries
 |      in :attr:`grad_input` and :attr:`grad_output` will be ``None`` for all non-Tensor
 |      arguments.
 |      
 |      For technical reasons, when this hook is applied to a Module, its forward function will
 |      receive a view of each Tensor passed to the Module. Similarly the caller will receive a view
 |      of each Tensor returned by the Module's forward function.
 |      
 |      .. warning ::
 |          Modifying inputs or outputs inplace is not allowed when using backward hooks and
 |          will raise an error.
 |      
 |      Args:
 |          hook (Callable): The user-defined hook to be registered.
 |          prepend (bool): If true, the provided ``hook`` will be fired before
 |              all existing ``backward`` hooks on this
 |              :class:`torch.nn.modules.Module`. Otherwise, the provided
 |              ``hook`` will be fired after all existing ``backward`` hooks on
 |              this :class:`torch.nn.modules.Module`. Note that global
 |              ``backward`` hooks registered with
 |              :func:`register_module_full_backward_hook` will fire before
 |              all hooks registered by this method.
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_full_backward_pre_hook(self, hook: Callable[[ForwardRef('Module'), Union[Tuple[torch.Tensor, ...], torch.Tensor]], Union[NoneType, Tuple[torch.Tensor, ...], torch.Tensor]], prepend: bool = False) -> torch.utils.hooks.RemovableHandle
 |      Registers a backward pre-hook on the module.
 |      
 |      The hook will be called every time the gradients for the module are computed.
 |      The hook should have the following signature::
 |      
 |          hook(module, grad_output) -> Tensor or None
 |      
 |      The :attr:`grad_output` is a tuple. The hook should
 |      not modify its arguments, but it can optionally return a new gradient with
 |      respect to the output that will be used in place of :attr:`grad_output` in
 |      subsequent computations. Entries in :attr:`grad_output` will be ``None`` for
 |      all non-Tensor arguments.
 |      
 |      For technical reasons, when this hook is applied to a Module, its forward function will
 |      receive a view of each Tensor passed to the Module. Similarly the caller will receive a view
 |      of each Tensor returned by the Module's forward function.
 |      
 |      .. warning ::
 |          Modifying inputs inplace is not allowed when using backward hooks and
 |          will raise an error.
 |      
 |      Args:
 |          hook (Callable): The user-defined hook to be registered.
 |          prepend (bool): If true, the provided ``hook`` will be fired before
 |              all existing ``backward_pre`` hooks on this
 |              :class:`torch.nn.modules.Module`. Otherwise, the provided
 |              ``hook`` will be fired after all existing ``backward_pre`` hooks
 |              on this :class:`torch.nn.modules.Module`. Note that global
 |              ``backward_pre`` hooks registered with
 |              :func:`register_module_full_backward_pre_hook` will fire before
 |              all hooks registered by this method.
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_load_state_dict_post_hook(self, hook)
 |      Registers a post hook to be run after module's ``load_state_dict``
 |      is called.
 |      
 |      It should have the following signature::
 |          hook(module, incompatible_keys) -> None
 |      
 |      The ``module`` argument is the current module that this hook is registered
 |      on, and the ``incompatible_keys`` argument is a ``NamedTuple`` consisting
 |      of attributes ``missing_keys`` and ``unexpected_keys``. ``missing_keys``
 |      is a ``list`` of ``str`` containing the missing keys and
 |      ``unexpected_keys`` is a ``list`` of ``str`` containing the unexpected keys.
 |      
 |      The given incompatible_keys can be modified inplace if needed.
 |      
 |      Note that the checks performed when calling :func:`load_state_dict` with
 |      ``strict=True`` are affected by modifications the hook makes to
 |      ``missing_keys`` or ``unexpected_keys``, as expected. Additions to either
 |      set of keys will result in an error being thrown when ``strict=True``, and
 |      clearing out both missing and unexpected keys will avoid an error.
 |      
 |      Returns:
 |          :class:`torch.utils.hooks.RemovableHandle`:
 |              a handle that can be used to remove the added hook by calling
 |              ``handle.remove()``
 |  
 |  register_module(self, name: str, module: Union[ForwardRef('Module'), NoneType]) -> None
 |      Alias for :func:`add_module`.
 |  
 |  register_parameter(self, name: str, param: Union[torch.nn.parameter.Parameter, NoneType]) -> None
 |      Adds a parameter to the module.
 |      
 |      The parameter can be accessed as an attribute using given name.
 |      
 |      Args:
 |          name (str): name of the parameter. The parameter can be accessed
 |              from this module using the given name
 |          param (Parameter or None): parameter to be added to the module. If
 |              ``None``, then operations that run on parameters, such as :attr:`cuda`,
 |              are ignored. If ``None``, the parameter is **not** included in the
 |              module's :attr:`state_dict`.
 |  
 |  register_state_dict_pre_hook(self, hook)
 |      These hooks will be called with arguments: ``self``, ``prefix``,
 |      and ``keep_vars`` before calling ``state_dict`` on ``self``. The registered
 |      hooks can be used to perform pre-processing before the ``state_dict``
 |      call is made.
 |  
 |  requires_grad_(self: ~T, requires_grad: bool = True) -> ~T
 |      Change if autograd should record operations on parameters in this
 |      module.
 |      
 |      This method sets the parameters' :attr:`requires_grad` attributes
 |      in-place.
 |      
 |      This method is helpful for freezing part of the module for finetuning
 |      or training parts of a model individually (e.g., GAN training).
 |      
 |      See :ref:`locally-disable-grad-doc` for a comparison between
 |      `.requires_grad_()` and several similar mechanisms that may be confused with it.
 |      
 |      Args:
 |          requires_grad (bool): whether autograd should record operations on
 |                                parameters in this module. Default: ``True``.
 |      
 |      Returns:
 |          Module: self
 |  
 |  set_extra_state(self, state: Any)
 |      This function is called from :func:`load_state_dict` to handle any extra state
 |      found within the `state_dict`. Implement this function and a corresponding
 |      :func:`get_extra_state` for your module if you need to store extra state within its
 |      `state_dict`.
 |      
 |      Args:
 |          state (dict): Extra state from the `state_dict`
 |  
 |  share_memory(self: ~T) -> ~T
 |      See :meth:`torch.Tensor.share_memory_`
 |  
 |  state_dict(self, *args, destination=None, prefix='', keep_vars=False)
 |      Returns a dictionary containing references to the whole state of the module.
 |      
 |      Both parameters and persistent buffers (e.g. running averages) are
 |      included. Keys are corresponding parameter and buffer names.
 |      Parameters and buffers set to ``None`` are not included.
 |      
 |      .. note::
 |          The returned object is a shallow copy. It contains references
 |          to the module's parameters and buffers.
 |      
 |      .. warning::
 |          Currently ``state_dict()`` also accepts positional arguments for
 |          ``destination``, ``prefix`` and ``keep_vars`` in order. However,
 |          this is being deprecated and keyword arguments will be enforced in
 |          future releases.
 |      
 |      .. warning::
 |          Please avoid the use of argument ``destination`` as it is not
 |          designed for end-users.
 |      
 |      Args:
 |          destination (dict, optional): If provided, the state of module will
 |              be updated into the dict and the same object is returned.
 |              Otherwise, an ``OrderedDict`` will be created and returned.
 |              Default: ``None``.
 |          prefix (str, optional): a prefix added to parameter and buffer
 |              names to compose the keys in state_dict. Default: ``''``.
 |          keep_vars (bool, optional): by default the :class:`~torch.Tensor` s
 |              returned in the state dict are detached from autograd. If it's
 |              set to ``True``, detaching will not be performed.
 |              Default: ``False``.
 |      
 |      Returns:
 |          dict:
 |              a dictionary containing a whole state of the module
 |      
 |      Example::
 |      
 |          >>> # xdoctest: +SKIP("undefined vars")
 |          >>> module.state_dict().keys()
 |          ['bias', 'weight']
 |  
 |  to(self, *args, **kwargs)
 |      Moves and/or casts the parameters and buffers.
 |      
 |      This can be called as
 |      
 |      .. function:: to(device=None, dtype=None, non_blocking=False)
 |         :noindex:
 |      
 |      .. function:: to(dtype, non_blocking=False)
 |         :noindex:
 |      
 |      .. function:: to(tensor, non_blocking=False)
 |         :noindex:
 |      
 |      .. function:: to(memory_format=torch.channels_last)
 |         :noindex:
 |      
 |      Its signature is similar to :meth:`torch.Tensor.to`, but only accepts
 |      floating point or complex :attr:`dtype`\ s. In addition, this method will
 |      only cast the floating point or complex parameters and buffers to :attr:`dtype`
 |      (if given). The integral parameters and buffers will be moved
 |      :attr:`device`, if that is given, but with dtypes unchanged. When
 |      :attr:`non_blocking` is set, it tries to convert/move asynchronously
 |      with respect to the host if possible, e.g., moving CPU Tensors with
 |      pinned memory to CUDA devices.
 |      
 |      See below for examples.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Args:
 |          device (:class:`torch.device`): the desired device of the parameters
 |              and buffers in this module
 |          dtype (:class:`torch.dtype`): the desired floating point or complex dtype of
 |              the parameters and buffers in this module
 |          tensor (torch.Tensor): Tensor whose dtype and device are the desired
 |              dtype and device for all parameters and buffers in this module
 |          memory_format (:class:`torch.memory_format`): the desired memory
 |              format for 4D parameters and buffers in this module (keyword
 |              only argument)
 |      
 |      Returns:
 |          Module: self
 |      
 |      Examples::
 |      
 |          >>> # xdoctest: +IGNORE_WANT("non-deterministic")
 |          >>> linear = nn.Linear(2, 2)
 |          >>> linear.weight
 |          Parameter containing:
 |          tensor([[ 0.1913, -0.3420],
 |                  [-0.5113, -0.2325]])
 |          >>> linear.to(torch.double)
 |          Linear(in_features=2, out_features=2, bias=True)
 |          >>> linear.weight
 |          Parameter containing:
 |          tensor([[ 0.1913, -0.3420],
 |                  [-0.5113, -0.2325]], dtype=torch.float64)
 |          >>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)
 |          >>> gpu1 = torch.device("cuda:1")
 |          >>> linear.to(gpu1, dtype=torch.half, non_blocking=True)
 |          Linear(in_features=2, out_features=2, bias=True)
 |          >>> linear.weight
 |          Parameter containing:
 |          tensor([[ 0.1914, -0.3420],
 |                  [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')
 |          >>> cpu = torch.device("cpu")
 |          >>> linear.to(cpu)
 |          Linear(in_features=2, out_features=2, bias=True)
 |          >>> linear.weight
 |          Parameter containing:
 |          tensor([[ 0.1914, -0.3420],
 |                  [-0.5112, -0.2324]], dtype=torch.float16)
 |      
 |          >>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)
 |          >>> linear.weight
 |          Parameter containing:
 |          tensor([[ 0.3741+0.j,  0.2382+0.j],
 |                  [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)
 |          >>> linear(torch.ones(3, 2, dtype=torch.cdouble))
 |          tensor([[0.6122+0.j, 0.1150+0.j],
 |                  [0.6122+0.j, 0.1150+0.j],
 |                  [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)
 |  
 |  to_empty(self: ~T, *, device: Union[str, torch.device]) -> ~T
 |      Moves the parameters and buffers to the specified device without copying storage.
 |      
 |      Args:
 |          device (:class:`torch.device`): The desired device of the parameters
 |              and buffers in this module.
 |      
 |      Returns:
 |          Module: self
 |  
 |  train(self: ~T, mode: bool = True) -> ~T
 |      Sets the module in training mode.
 |      
 |      This has any effect only on certain modules. See documentations of
 |      particular modules for details of their behaviors in training/evaluation
 |      mode, if they are affected, e.g. :class:`Dropout`, :class:`BatchNorm`,
 |      etc.
 |      
 |      Args:
 |          mode (bool): whether to set training mode (``True``) or evaluation
 |                       mode (``False``). Default: ``True``.
 |      
 |      Returns:
 |          Module: self
 |  
 |  type(self: ~T, dst_type: Union[torch.dtype, str]) -> ~T
 |      Casts all parameters and buffers to :attr:`dst_type`.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Args:
 |          dst_type (type or string): the desired type
 |      
 |      Returns:
 |          Module: self
 |  
 |  xpu(self: ~T, device: Union[int, torch.device, NoneType] = None) -> ~T
 |      Moves all model parameters and buffers to the XPU.
 |      
 |      This also makes associated parameters and buffers different objects. So
 |      it should be called before constructing optimizer if the module will
 |      live on XPU while being optimized.
 |      
 |      .. note::
 |          This method modifies the module in-place.
 |      
 |      Arguments:
 |          device (int, optional): if specified, all parameters will be
 |              copied to that device
 |      
 |      Returns:
 |          Module: self
 |  
 |  zero_grad(self, set_to_none: bool = True) -> None
 |      Sets gradients of all model parameters to zero. See similar function
 |      under :class:`torch.optim.Optimizer` for more context.
 |      
 |      Args:
 |          set_to_none (bool): instead of setting to zero, set the grads to None.
 |              See :meth:`torch.optim.Optimizer.zero_grad` for details.
 |  
 |  ----------------------------------------------------------------------
 |  Data descriptors inherited from torch.nn.modules.module.Module:
 |  
 |  __dict__
 |      dictionary for instance variables (if defined)
 |  
 |  __weakref__
 |      list of weak references to the object (if defined)
 |  
 |  ----------------------------------------------------------------------
 |  Data and other attributes inherited from torch.nn.modules.module.Module:
 |  
 |  T_destination = ~T_destination
 |  
 |  __annotations__ = {'__call__': typing.Callable[..., typing.Any], '_bac...
 |  
 |  call_super_init = False
 |  
 |  dump_patches = False

def look_ahead_iterator(gen):
    prev = None
    for item in gen:
        if prev is not None:
            yield (prev, item)
        prev = item

class Bigrams(IterableDataset):
  def __init__(self, text_file, vocabulary_size):
      self.vocab = build_vocab_from_iterator(
         get_word_lines_from_file(text_file),
         max_tokens = vocabulary_size,
         specials = ['<unk>'])
      self.vocab.set_default_index(self.vocab['<unk>'])
      self.vocabulary_size = vocabulary_size
      self.text_file = text_file

  def __iter__(self):
     return look_ahead_iterator(
         (self.vocab[t] for t in itertools.chain.from_iterable(get_word_lines_from_file(self.text_file))))

train_dataset = Bigrams(train_file, vocab_size)

print(train_dataset)

<__main__.Bigrams object at 0x7fdd26d23940>

torch.cuda.memory_summary(device=None, abbreviated=False)

'|===========================================================================|\n|                  PyTorch CUDA memory summary, device ID 0                 |\n|---------------------------------------------------------------------------|\n|            CUDA OOMs: 1            |        cudaMalloc retries: 1         |\n|===========================================================================|\n|        Metric         | Cur Usage  | Peak Usage | Tot Alloc  | Tot Freed  |\n|---------------------------------------------------------------------------|\n| Allocated memory      | 699613 KiB |   1903 MiB |   3735 MiB |   3052 MiB |\n|       from large pool | 699414 KiB |   1903 MiB |   3734 MiB |   3051 MiB |\n|       from small pool |    199 KiB |      1 MiB |      1 MiB |      1 MiB |\n|---------------------------------------------------------------------------|\n| Active memory         | 699613 KiB |   1903 MiB |   3735 MiB |   3052 MiB |\n|       from large pool | 699414 KiB |   1903 MiB |   3734 MiB |   3051 MiB |\n|       from small pool |    199 KiB |      1 MiB |      1 MiB |      1 MiB |\n|---------------------------------------------------------------------------|\n| Requested memory      | 699611 KiB |   1903 MiB |   3735 MiB |   3052 MiB |\n|       from large pool | 699413 KiB |   1903 MiB |   3734 MiB |   3051 MiB |\n|       from small pool |    197 KiB |      1 MiB |      1 MiB |      1 MiB |\n|---------------------------------------------------------------------------|\n| GPU reserved memory   | 710656 KiB |   1918 MiB |   1918 MiB |   1224 MiB |\n|       from large pool | 708608 KiB |   1916 MiB |   1916 MiB |   1224 MiB |\n|       from small pool |   2048 KiB |      2 MiB |      2 MiB |      0 MiB |\n|---------------------------------------------------------------------------|\n| Non-releasable memory |  11043 KiB |  19364 KiB |  28939 KiB |  17896 KiB |\n|       from large pool |   9194 KiB |  17514 KiB |  25954 KiB |  16760 KiB |\n|       from small pool |   1849 KiB |   1950 KiB |   2985 KiB |   1136 KiB |\n|---------------------------------------------------------------------------|\n| Allocations           |      10    |      17    |      38    |      28    |\n|       from large pool |       5    |       7    |      10    |       5    |\n|       from small pool |       5    |      11    |      28    |      23    |\n|---------------------------------------------------------------------------|\n| Active allocs         |      10    |      17    |      38    |      28    |\n|       from large pool |       5    |       7    |      10    |       5    |\n|       from small pool |       5    |      11    |      28    |      23    |\n|---------------------------------------------------------------------------|\n| GPU reserved segments |       5    |       7    |       7    |       2    |\n|       from large pool |       4    |       6    |       6    |       2    |\n|       from small pool |       1    |       1    |       1    |       0    |\n|---------------------------------------------------------------------------|\n| Non-releasable allocs |       6    |       8    |      20    |      14    |\n|       from large pool |       4    |       6    |       9    |       5    |\n|       from small pool |       2    |       3    |      11    |       9    |\n|---------------------------------------------------------------------------|\n| Oversize allocations  |       0    |       0    |       0    |       0    |\n|---------------------------------------------------------------------------|\n| Oversize GPU segments |       0    |       0    |       0    |       0    |\n|===========================================================================|\n'

import os
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "max_split_size_mb:256"

device = 'cuda'
model = SimpleBigramNeuralLanguageModel(vocab_size, embed_size).to(device)

data = DataLoader(train_dataset, batch_size=batch_s)
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
criterion = torch.nn.NLLLoss()
torch.cuda.empty_cache()
gc.collect()

model.load_state_dict(torch.load('model-bigram_final.bin'))
for i in range(1, epochs+1):
    print('epoch: =', i)
    model.train()
    step = 0
    for x, y in data: # prev, predicting, following words
       x = x.to(device)
       y = y.to(device)
       optimizer.zero_grad()
       ypredicted = model(x) #previous, following word
       loss = criterion(torch.log(ypredicted), y)
       if step % 100 == 0:
          print(step, loss)
       step += 1
       loss.backward()
       optimizer.step()
    torch.save(model.state_dict(), f'model-bigram_2nd-run{i}.bin')    
torch.save(model.state_dict(), f'model-bigram_final.bin')

epoch: = 1

/home/gedin/.local/lib/python3.8/site-packages/torch/nn/modules/container.py:217: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  input = module(input)

0 tensor(5.9599, device='cuda:0', grad_fn=<NllLossBackward0>)
1000
100 tensor(6.1015, device='cuda:0', grad_fn=<NllLossBackward0>)
200 tensor(5.9708, device='cuda:0', grad_fn=<NllLossBackward0>)
2000
300 tensor(6.2176, device='cuda:0', grad_fn=<NllLossBackward0>)
3000
400 tensor(5.9401, device='cuda:0', grad_fn=<NllLossBackward0>)
4000
500 tensor(6.2084, device='cuda:0', grad_fn=<NllLossBackward0>)
5000
600 tensor(5.9736, device='cuda:0', grad_fn=<NllLossBackward0>)
6000
700 tensor(6.1423, device='cuda:0', grad_fn=<NllLossBackward0>)
7000
800 tensor(5.7344, device='cuda:0', grad_fn=<NllLossBackward0>)
8000
900 tensor(6.0950, device='cuda:0', grad_fn=<NllLossBackward0>)
9000
1000 tensor(5.8473, device='cuda:0', grad_fn=<NllLossBackward0>)
10000
1100 tensor(6.0612, device='cuda:0', grad_fn=<NllLossBackward0>)
11000
1200 tensor(6.1509, device='cuda:0', grad_fn=<NllLossBackward0>)
12000
1300 tensor(6.0760, device='cuda:0', grad_fn=<NllLossBackward0>)
13000
1400 tensor(6.2047, device='cuda:0', grad_fn=<NllLossBackward0>)
14000
1500 tensor(6.1186, device='cuda:0', grad_fn=<NllLossBackward0>)
15000
1600 tensor(5.8722, device='cuda:0', grad_fn=<NllLossBackward0>)
16000
1700 tensor(5.8741, device='cuda:0', grad_fn=<NllLossBackward0>)
17000
1800 tensor(5.8971, device='cuda:0', grad_fn=<NllLossBackward0>)
18000
1900 tensor(5.8521, device='cuda:0', grad_fn=<NllLossBackward0>)
19000
2000 tensor(5.9434, device='cuda:0', grad_fn=<NllLossBackward0>)
20000
2100 tensor(6.0348, device='cuda:0', grad_fn=<NllLossBackward0>)
21000
2200 tensor(5.8840, device='cuda:0', grad_fn=<NllLossBackward0>)
22000
2300 tensor(5.8641, device='cuda:0', grad_fn=<NllLossBackward0>)
23000
2400 tensor(5.9068, device='cuda:0', grad_fn=<NllLossBackward0>)
24000
2500 tensor(5.9170, device='cuda:0', grad_fn=<NllLossBackward0>)
25000
2600 tensor(5.9812, device='cuda:0', grad_fn=<NllLossBackward0>)
26000
2700 tensor(5.8985, device='cuda:0', grad_fn=<NllLossBackward0>)
27000
2800 tensor(6.0008, device='cuda:0', grad_fn=<NllLossBackward0>)
28000
2900 tensor(6.1230, device='cuda:0', grad_fn=<NllLossBackward0>)
29000
3000 tensor(5.8770, device='cuda:0', grad_fn=<NllLossBackward0>)
30000
3100 tensor(5.9268, device='cuda:0', grad_fn=<NllLossBackward0>)
31000
3200 tensor(5.8530, device='cuda:0', grad_fn=<NllLossBackward0>)
32000
3300 tensor(5.8436, device='cuda:0', grad_fn=<NllLossBackward0>)
33000
3400 tensor(5.7692, device='cuda:0', grad_fn=<NllLossBackward0>)
34000
3500 tensor(5.8909, device='cuda:0', grad_fn=<NllLossBackward0>)
35000
3600 tensor(5.8325, device='cuda:0', grad_fn=<NllLossBackward0>)
36000
3700 tensor(5.8082, device='cuda:0', grad_fn=<NllLossBackward0>)
37000
3800 tensor(5.8106, device='cuda:0', grad_fn=<NllLossBackward0>)
38000
3900 tensor(5.6382, device='cuda:0', grad_fn=<NllLossBackward0>)
39000
4000 tensor(5.6596, device='cuda:0', grad_fn=<NllLossBackward0>)
40000
4100 tensor(5.9587, device='cuda:0', grad_fn=<NllLossBackward0>)
41000
4200 tensor(5.8862, device='cuda:0', grad_fn=<NllLossBackward0>)
42000
4300 tensor(5.9541, device='cuda:0', grad_fn=<NllLossBackward0>)
43000
4400 tensor(5.8681, device='cuda:0', grad_fn=<NllLossBackward0>)
44000
4500 tensor(5.6963, device='cuda:0', grad_fn=<NllLossBackward0>)
45000
4600 tensor(6.0707, device='cuda:0', grad_fn=<NllLossBackward0>)
46000
4700 tensor(5.7091, device='cuda:0', grad_fn=<NllLossBackward0>)
47000
4800 tensor(5.8139, device='cuda:0', grad_fn=<NllLossBackward0>)
48000
4900 tensor(5.8696, device='cuda:0', grad_fn=<NllLossBackward0>)
49000
5000 tensor(5.8844, device='cuda:0', grad_fn=<NllLossBackward0>)
50000
5100 tensor(5.9806, device='cuda:0', grad_fn=<NllLossBackward0>)
51000
5200 tensor(6.0075, device='cuda:0', grad_fn=<NllLossBackward0>)
52000
5300 tensor(6.0588, device='cuda:0', grad_fn=<NllLossBackward0>)
53000
5400 tensor(5.8456, device='cuda:0', grad_fn=<NllLossBackward0>)
54000
5500 tensor(5.9166, device='cuda:0', grad_fn=<NllLossBackward0>)
55000
5600 tensor(5.6528, device='cuda:0', grad_fn=<NllLossBackward0>)
56000
5700 tensor(5.8988, device='cuda:0', grad_fn=<NllLossBackward0>)
57000
5800 tensor(5.9132, device='cuda:0', grad_fn=<NllLossBackward0>)
58000
5900 tensor(5.9460, device='cuda:0', grad_fn=<NllLossBackward0>)
59000
6000 tensor(5.7543, device='cuda:0', grad_fn=<NllLossBackward0>)
60000
6100 tensor(5.8256, device='cuda:0', grad_fn=<NllLossBackward0>)
61000
6200 tensor(5.9448, device='cuda:0', grad_fn=<NllLossBackward0>)
62000
6300 tensor(5.7601, device='cuda:0', grad_fn=<NllLossBackward0>)
63000
6400 tensor(5.7091, device='cuda:0', grad_fn=<NllLossBackward0>)
64000
6500 tensor(5.5621, device='cuda:0', grad_fn=<NllLossBackward0>)
65000
6600 tensor(5.7094, device='cuda:0', grad_fn=<NllLossBackward0>)
66000
6700 tensor(5.6785, device='cuda:0', grad_fn=<NllLossBackward0>)
67000
6800 tensor(5.9249, device='cuda:0', grad_fn=<NllLossBackward0>)
68000
6900 tensor(5.8775, device='cuda:0', grad_fn=<NllLossBackward0>)
69000
7000 tensor(5.8075, device='cuda:0', grad_fn=<NllLossBackward0>)
70000
7100 tensor(5.5748, device='cuda:0', grad_fn=<NllLossBackward0>)
71000
7200 tensor(5.7217, device='cuda:0', grad_fn=<NllLossBackward0>)
72000
7300 tensor(5.9124, device='cuda:0', grad_fn=<NllLossBackward0>)
73000
7400 tensor(5.7197, device='cuda:0', grad_fn=<NllLossBackward0>)
74000
7500 tensor(5.6429, device='cuda:0', grad_fn=<NllLossBackward0>)
75000
7600 tensor(5.6847, device='cuda:0', grad_fn=<NllLossBackward0>)
76000
7700 tensor(5.7197, device='cuda:0', grad_fn=<NllLossBackward0>)
77000
7800 tensor(5.8559, device='cuda:0', grad_fn=<NllLossBackward0>)
78000
7900 tensor(5.5600, device='cuda:0', grad_fn=<NllLossBackward0>)
79000
8000 tensor(5.6288, device='cuda:0', grad_fn=<NllLossBackward0>)
80000
8100 tensor(5.7767, device='cuda:0', grad_fn=<NllLossBackward0>)
81000
8200 tensor(5.8037, device='cuda:0', grad_fn=<NllLossBackward0>)
82000
8300 tensor(5.7344, device='cuda:0', grad_fn=<NllLossBackward0>)
83000
8400 tensor(5.8092, device='cuda:0', grad_fn=<NllLossBackward0>)
84000
8500 tensor(5.8847, device='cuda:0', grad_fn=<NllLossBackward0>)
85000
8600 tensor(5.8754, device='cuda:0', grad_fn=<NllLossBackward0>)
86000
8700 tensor(5.9227, device='cuda:0', grad_fn=<NllLossBackward0>)
87000
8800 tensor(5.8028, device='cuda:0', grad_fn=<NllLossBackward0>)
88000
8900 tensor(5.6476, device='cuda:0', grad_fn=<NllLossBackward0>)
89000
9000 tensor(5.7656, device='cuda:0', grad_fn=<NllLossBackward0>)
90000
9100 tensor(5.7805, device='cuda:0', grad_fn=<NllLossBackward0>)
91000
9200 tensor(5.6879, device='cuda:0', grad_fn=<NllLossBackward0>)
92000
9300 tensor(5.7098, device='cuda:0', grad_fn=<NllLossBackward0>)
93000
9400 tensor(5.5631, device='cuda:0', grad_fn=<NllLossBackward0>)
94000
9500 tensor(5.6497, device='cuda:0', grad_fn=<NllLossBackward0>)
95000
9600 tensor(5.7500, device='cuda:0', grad_fn=<NllLossBackward0>)
96000
9700 tensor(5.6607, device='cuda:0', grad_fn=<NllLossBackward0>)
97000
9800 tensor(5.7196, device='cuda:0', grad_fn=<NllLossBackward0>)
9900 tensor(5.5987, device='cuda:0', grad_fn=<NllLossBackward0>)
98000
10000 tensor(5.7795, device='cuda:0', grad_fn=<NllLossBackward0>)
99000
10100 tensor(5.6980, device='cuda:0', grad_fn=<NllLossBackward0>)
100000
10200 tensor(5.6093, device='cuda:0', grad_fn=<NllLossBackward0>)
101000
10300 tensor(5.6792, device='cuda:0', grad_fn=<NllLossBackward0>)
102000
10400 tensor(5.7035, device='cuda:0', grad_fn=<NllLossBackward0>)
103000
10500 tensor(5.8282, device='cuda:0', grad_fn=<NllLossBackward0>)
104000
10600 tensor(5.8605, device='cuda:0', grad_fn=<NllLossBackward0>)
105000
10700 tensor(5.7354, device='cuda:0', grad_fn=<NllLossBackward0>)
106000
10800 tensor(5.8034, device='cuda:0', grad_fn=<NllLossBackward0>)
107000
10900 tensor(5.6194, device='cuda:0', grad_fn=<NllLossBackward0>)
108000
11000 tensor(5.8502, device='cuda:0', grad_fn=<NllLossBackward0>)
109000
11100 tensor(5.4406, device='cuda:0', grad_fn=<NllLossBackward0>)
110000
11200 tensor(5.6379, device='cuda:0', grad_fn=<NllLossBackward0>)
111000
11300 tensor(5.6668, device='cuda:0', grad_fn=<NllLossBackward0>)
112000
11400 tensor(5.6140, device='cuda:0', grad_fn=<NllLossBackward0>)
113000
11500 tensor(5.6565, device='cuda:0', grad_fn=<NllLossBackward0>)
114000
11600 tensor(5.6308, device='cuda:0', grad_fn=<NllLossBackward0>)
115000
11700 tensor(5.5680, device='cuda:0', grad_fn=<NllLossBackward0>)
116000
11800 tensor(5.7604, device='cuda:0', grad_fn=<NllLossBackward0>)
117000
11900 tensor(5.5792, device='cuda:0', grad_fn=<NllLossBackward0>)
118000
12000 tensor(5.7329, device='cuda:0', grad_fn=<NllLossBackward0>)
119000
12100 tensor(5.7726, device='cuda:0', grad_fn=<NllLossBackward0>)
120000
12200 tensor(5.7151, device='cuda:0', grad_fn=<NllLossBackward0>)
121000
12300 tensor(5.8561, device='cuda:0', grad_fn=<NllLossBackward0>)
122000
12400 tensor(5.6791, device='cuda:0', grad_fn=<NllLossBackward0>)
123000
12500 tensor(5.5574, device='cuda:0', grad_fn=<NllLossBackward0>)
124000
12600 tensor(5.6817, device='cuda:0', grad_fn=<NllLossBackward0>)
125000
12700 tensor(5.5375, device='cuda:0', grad_fn=<NllLossBackward0>)
126000
12800 tensor(5.7270, device='cuda:0', grad_fn=<NllLossBackward0>)
127000
12900 tensor(5.6252, device='cuda:0', grad_fn=<NllLossBackward0>)
128000
13000 tensor(5.4536, device='cuda:0', grad_fn=<NllLossBackward0>)
129000
13100 tensor(5.6091, device='cuda:0', grad_fn=<NllLossBackward0>)
130000
13200 tensor(5.7324, device='cuda:0', grad_fn=<NllLossBackward0>)
131000
13300 tensor(5.5253, device='cuda:0', grad_fn=<NllLossBackward0>)
132000
13400 tensor(5.6491, device='cuda:0', grad_fn=<NllLossBackward0>)
133000
13500 tensor(5.5728, device='cuda:0', grad_fn=<NllLossBackward0>)
134000
13600 tensor(5.6632, device='cuda:0', grad_fn=<NllLossBackward0>)
135000
13700 tensor(5.6678, device='cuda:0', grad_fn=<NllLossBackward0>)
136000
13800 tensor(5.6112, device='cuda:0', grad_fn=<NllLossBackward0>)
137000
13900 tensor(5.4884, device='cuda:0', grad_fn=<NllLossBackward0>)
138000
14000 tensor(5.7304, device='cuda:0', grad_fn=<NllLossBackward0>)
139000
14100 tensor(5.4326, device='cuda:0', grad_fn=<NllLossBackward0>)
140000
14200 tensor(5.7188, device='cuda:0', grad_fn=<NllLossBackward0>)
141000
14300 tensor(5.6519, device='cuda:0', grad_fn=<NllLossBackward0>)
142000
14400 tensor(5.5892, device='cuda:0', grad_fn=<NllLossBackward0>)
143000
14500 tensor(5.7225, device='cuda:0', grad_fn=<NllLossBackward0>)
144000
14600 tensor(5.7216, device='cuda:0', grad_fn=<NllLossBackward0>)
145000
14700 tensor(5.5748, device='cuda:0', grad_fn=<NllLossBackward0>)
146000
14800 tensor(6.0184, device='cuda:0', grad_fn=<NllLossBackward0>)
147000
14900 tensor(5.6781, device='cuda:0', grad_fn=<NllLossBackward0>)
148000
15000 tensor(5.6038, device='cuda:0', grad_fn=<NllLossBackward0>)
149000
15100 tensor(5.7875, device='cuda:0', grad_fn=<NllLossBackward0>)
150000
15200 tensor(5.6485, device='cuda:0', grad_fn=<NllLossBackward0>)
151000
15300 tensor(5.5927, device='cuda:0', grad_fn=<NllLossBackward0>)
152000
15400 tensor(5.5156, device='cuda:0', grad_fn=<NllLossBackward0>)
153000
15500 tensor(5.6556, device='cuda:0', grad_fn=<NllLossBackward0>)
154000
15600 tensor(5.6485, device='cuda:0', grad_fn=<NllLossBackward0>)
155000
15700 tensor(5.5904, device='cuda:0', grad_fn=<NllLossBackward0>)
156000
15800 tensor(5.4613, device='cuda:0', grad_fn=<NllLossBackward0>)
157000
15900 tensor(5.6254, device='cuda:0', grad_fn=<NllLossBackward0>)
158000
16000 tensor(5.4349, device='cuda:0', grad_fn=<NllLossBackward0>)
159000
16100 tensor(5.5205, device='cuda:0', grad_fn=<NllLossBackward0>)
160000
16200 tensor(5.8051, device='cuda:0', grad_fn=<NllLossBackward0>)
161000
16300 tensor(5.6452, device='cuda:0', grad_fn=<NllLossBackward0>)
162000
16400 tensor(5.6071, device='cuda:0', grad_fn=<NllLossBackward0>)
163000
16500 tensor(5.7237, device='cuda:0', grad_fn=<NllLossBackward0>)
164000
16600 tensor(5.5771, device='cuda:0', grad_fn=<NllLossBackward0>)
165000
16700 tensor(5.5355, device='cuda:0', grad_fn=<NllLossBackward0>)
166000
16800 tensor(5.6363, device='cuda:0', grad_fn=<NllLossBackward0>)
167000
16900 tensor(5.3746, device='cuda:0', grad_fn=<NllLossBackward0>)
168000
17000 tensor(5.6707, device='cuda:0', grad_fn=<NllLossBackward0>)
169000
17100 tensor(5.5359, device='cuda:0', grad_fn=<NllLossBackward0>)
170000
17200 tensor(5.6118, device='cuda:0', grad_fn=<NllLossBackward0>)
171000
17300 tensor(5.6740, device='cuda:0', grad_fn=<NllLossBackward0>)
172000
17400 tensor(5.4438, device='cuda:0', grad_fn=<NllLossBackward0>)
173000
17500 tensor(5.5001, device='cuda:0', grad_fn=<NllLossBackward0>)
174000
17600 tensor(5.4953, device='cuda:0', grad_fn=<NllLossBackward0>)
175000
17700 tensor(5.5398, device='cuda:0', grad_fn=<NllLossBackward0>)
176000
17800 tensor(5.6053, device='cuda:0', grad_fn=<NllLossBackward0>)
177000
17900 tensor(5.4726, device='cuda:0', grad_fn=<NllLossBackward0>)
178000
18000 tensor(5.6747, device='cuda:0', grad_fn=<NllLossBackward0>)
179000
18100 tensor(5.6238, device='cuda:0', grad_fn=<NllLossBackward0>)
180000
18200 tensor(5.5469, device='cuda:0', grad_fn=<NllLossBackward0>)
181000
18300 tensor(5.5299, device='cuda:0', grad_fn=<NllLossBackward0>)
182000
18400 tensor(5.6323, device='cuda:0', grad_fn=<NllLossBackward0>)
183000
18500 tensor(5.5893, device='cuda:0', grad_fn=<NllLossBackward0>)
184000
18600 tensor(5.7452, device='cuda:0', grad_fn=<NllLossBackward0>)
185000
18700 tensor(5.5576, device='cuda:0', grad_fn=<NllLossBackward0>)
186000
18800 tensor(5.7439, device='cuda:0', grad_fn=<NllLossBackward0>)
187000
18900 tensor(5.6106, device='cuda:0', grad_fn=<NllLossBackward0>)
188000
19000 tensor(5.6647, device='cuda:0', grad_fn=<NllLossBackward0>)
189000
19100 tensor(5.7728, device='cuda:0', grad_fn=<NllLossBackward0>)
190000
19200 tensor(5.6169, device='cuda:0', grad_fn=<NllLossBackward0>)
191000
19300 tensor(5.7852, device='cuda:0', grad_fn=<NllLossBackward0>)
192000
19400 tensor(5.5627, device='cuda:0', grad_fn=<NllLossBackward0>)
193000
19500 tensor(5.5682, device='cuda:0', grad_fn=<NllLossBackward0>)
194000
19600 tensor(5.5978, device='cuda:0', grad_fn=<NllLossBackward0>)
195000
19700 tensor(5.6453, device='cuda:0', grad_fn=<NllLossBackward0>)
196000
19800 tensor(5.4786, device='cuda:0', grad_fn=<NllLossBackward0>)
197000
19900 tensor(5.4894, device='cuda:0', grad_fn=<NllLossBackward0>)
198000
20000 tensor(5.4999, device='cuda:0', grad_fn=<NllLossBackward0>)
199000
20100 tensor(5.4881, device='cuda:0', grad_fn=<NllLossBackward0>)
200000
20200 tensor(5.3915, device='cuda:0', grad_fn=<NllLossBackward0>)
201000
20300 tensor(5.5216, device='cuda:0', grad_fn=<NllLossBackward0>)
20400 tensor(5.5761, device='cuda:0', grad_fn=<NllLossBackward0>)
202000
20500 tensor(5.5586, device='cuda:0', grad_fn=<NllLossBackward0>)
203000
20600 tensor(5.7870, device='cuda:0', grad_fn=<NllLossBackward0>)
204000
20700 tensor(5.5776, device='cuda:0', grad_fn=<NllLossBackward0>)
205000
20800 tensor(5.4417, device='cuda:0', grad_fn=<NllLossBackward0>)
206000
20900 tensor(5.7186, device='cuda:0', grad_fn=<NllLossBackward0>)
207000
21000 tensor(5.5415, device='cuda:0', grad_fn=<NllLossBackward0>)
208000
21100 tensor(5.5141, device='cuda:0', grad_fn=<NllLossBackward0>)
209000
21200 tensor(5.4401, device='cuda:0', grad_fn=<NllLossBackward0>)
210000
21300 tensor(5.6511, device='cuda:0', grad_fn=<NllLossBackward0>)
211000
21400 tensor(5.6474, device='cuda:0', grad_fn=<NllLossBackward0>)
212000
21500 tensor(5.3946, device='cuda:0', grad_fn=<NllLossBackward0>)
213000
21600 tensor(5.3958, device='cuda:0', grad_fn=<NllLossBackward0>)
214000
21700 tensor(5.4040, device='cuda:0', grad_fn=<NllLossBackward0>)
215000
21800 tensor(5.5745, device='cuda:0', grad_fn=<NllLossBackward0>)
216000
21900 tensor(5.4996, device='cuda:0', grad_fn=<NllLossBackward0>)
217000
22000 tensor(5.5234, device='cuda:0', grad_fn=<NllLossBackward0>)
218000
22100 tensor(5.3870, device='cuda:0', grad_fn=<NllLossBackward0>)
219000
22200 tensor(5.2661, device='cuda:0', grad_fn=<NllLossBackward0>)
220000
22300 tensor(5.7031, device='cuda:0', grad_fn=<NllLossBackward0>)
221000
22400 tensor(5.3633, device='cuda:0', grad_fn=<NllLossBackward0>)
222000
22500 tensor(5.4404, device='cuda:0', grad_fn=<NllLossBackward0>)
223000
22600 tensor(5.5951, device='cuda:0', grad_fn=<NllLossBackward0>)
224000
22700 tensor(5.3901, device='cuda:0', grad_fn=<NllLossBackward0>)
225000
22800 tensor(5.6404, device='cuda:0', grad_fn=<NllLossBackward0>)
226000
22900 tensor(5.6646, device='cuda:0', grad_fn=<NllLossBackward0>)
227000
23000 tensor(5.5949, device='cuda:0', grad_fn=<NllLossBackward0>)
228000
23100 tensor(5.5284, device='cuda:0', grad_fn=<NllLossBackward0>)
229000
23200 tensor(5.5617, device='cuda:0', grad_fn=<NllLossBackward0>)
230000
23300 tensor(5.6426, device='cuda:0', grad_fn=<NllLossBackward0>)
231000
23400 tensor(5.7283, device='cuda:0', grad_fn=<NllLossBackward0>)
232000
23500 tensor(5.4558, device='cuda:0', grad_fn=<NllLossBackward0>)
233000
23600 tensor(5.4600, device='cuda:0', grad_fn=<NllLossBackward0>)
234000
23700 tensor(5.4961, device='cuda:0', grad_fn=<NllLossBackward0>)
235000
23800 tensor(5.3373, device='cuda:0', grad_fn=<NllLossBackward0>)
236000
23900 tensor(5.4470, device='cuda:0', grad_fn=<NllLossBackward0>)
237000
24000 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>)
238000
24100 tensor(5.5112, device='cuda:0', grad_fn=<NllLossBackward0>)
239000
24200 tensor(5.6918, device='cuda:0', grad_fn=<NllLossBackward0>)
240000
24300 tensor(5.6115, device='cuda:0', grad_fn=<NllLossBackward0>)
241000
24400 tensor(5.7404, device='cuda:0', grad_fn=<NllLossBackward0>)
242000
24500 tensor(5.4982, device='cuda:0', grad_fn=<NllLossBackward0>)
243000
24600 tensor(5.6136, device='cuda:0', grad_fn=<NllLossBackward0>)
244000
24700 tensor(5.5225, device='cuda:0', grad_fn=<NllLossBackward0>)
245000
24800 tensor(5.5563, device='cuda:0', grad_fn=<NllLossBackward0>)
246000
24900 tensor(5.6283, device='cuda:0', grad_fn=<NllLossBackward0>)
247000
25000 tensor(5.6176, device='cuda:0', grad_fn=<NllLossBackward0>)
248000
25100 tensor(5.5795, device='cuda:0', grad_fn=<NllLossBackward0>)
249000
25200 tensor(5.5831, device='cuda:0', grad_fn=<NllLossBackward0>)
250000
25300 tensor(5.5894, device='cuda:0', grad_fn=<NllLossBackward0>)
251000
25400 tensor(5.5670, device='cuda:0', grad_fn=<NllLossBackward0>)
252000
25500 tensor(5.5016, device='cuda:0', grad_fn=<NllLossBackward0>)
253000
25600 tensor(5.7909, device='cuda:0', grad_fn=<NllLossBackward0>)
254000
25700 tensor(5.5229, device='cuda:0', grad_fn=<NllLossBackward0>)
255000
25800 tensor(5.6035, device='cuda:0', grad_fn=<NllLossBackward0>)
256000
25900 tensor(5.5293, device='cuda:0', grad_fn=<NllLossBackward0>)
257000
26000 tensor(5.5553, device='cuda:0', grad_fn=<NllLossBackward0>)
258000
26100 tensor(5.4476, device='cuda:0', grad_fn=<NllLossBackward0>)
259000
26200 tensor(5.3721, device='cuda:0', grad_fn=<NllLossBackward0>)
260000
26300 tensor(5.6142, device='cuda:0', grad_fn=<NllLossBackward0>)
261000
26400 tensor(5.6202, device='cuda:0', grad_fn=<NllLossBackward0>)
262000
26500 tensor(5.3529, device='cuda:0', grad_fn=<NllLossBackward0>)
263000
26600 tensor(5.7148, device='cuda:0', grad_fn=<NllLossBackward0>)
264000
26700 tensor(5.5755, device='cuda:0', grad_fn=<NllLossBackward0>)
265000
26800 tensor(5.7480, device='cuda:0', grad_fn=<NllLossBackward0>)
266000
26900 tensor(5.5025, device='cuda:0', grad_fn=<NllLossBackward0>)
267000
27000 tensor(5.4017, device='cuda:0', grad_fn=<NllLossBackward0>)
268000
27100 tensor(5.3996, device='cuda:0', grad_fn=<NllLossBackward0>)
269000
27200 tensor(5.4862, device='cuda:0', grad_fn=<NllLossBackward0>)
270000
27300 tensor(5.6392, device='cuda:0', grad_fn=<NllLossBackward0>)
271000
27400 tensor(5.5634, device='cuda:0', grad_fn=<NllLossBackward0>)
272000
27500 tensor(5.4420, device='cuda:0', grad_fn=<NllLossBackward0>)
273000
27600 tensor(5.7835, device='cuda:0', grad_fn=<NllLossBackward0>)
274000
27700 tensor(5.5555, device='cuda:0', grad_fn=<NllLossBackward0>)
275000
27800 tensor(5.5381, device='cuda:0', grad_fn=<NllLossBackward0>)
276000
27900 tensor(5.6515, device='cuda:0', grad_fn=<NllLossBackward0>)
277000
28000 tensor(5.5254, device='cuda:0', grad_fn=<NllLossBackward0>)
278000
28100 tensor(5.4929, device='cuda:0', grad_fn=<NllLossBackward0>)
279000
28200 tensor(5.6218, device='cuda:0', grad_fn=<NllLossBackward0>)
280000
28300 tensor(5.2878, device='cuda:0', grad_fn=<NllLossBackward0>)
281000
28400 tensor(5.7112, device='cuda:0', grad_fn=<NllLossBackward0>)
282000
28500 tensor(5.5490, device='cuda:0', grad_fn=<NllLossBackward0>)
283000
28600 tensor(5.4572, device='cuda:0', grad_fn=<NllLossBackward0>)
284000
28700 tensor(5.6349, device='cuda:0', grad_fn=<NllLossBackward0>)
285000
28800 tensor(5.6607, device='cuda:0', grad_fn=<NllLossBackward0>)
286000
28900 tensor(5.5422, device='cuda:0', grad_fn=<NllLossBackward0>)
287000
29000 tensor(5.4277, device='cuda:0', grad_fn=<NllLossBackward0>)
288000
29100 tensor(5.1870, device='cuda:0', grad_fn=<NllLossBackward0>)
289000
29200 tensor(5.3593, device='cuda:0', grad_fn=<NllLossBackward0>)
290000
29300 tensor(5.6512, device='cuda:0', grad_fn=<NllLossBackward0>)
291000
29400 tensor(5.8051, device='cuda:0', grad_fn=<NllLossBackward0>)
292000
29500 tensor(5.5308, device='cuda:0', grad_fn=<NllLossBackward0>)
293000
29600 tensor(5.3791, device='cuda:0', grad_fn=<NllLossBackward0>)
294000
29700 tensor(5.6108, device='cuda:0', grad_fn=<NllLossBackward0>)
295000
29800 tensor(5.4015, device='cuda:0', grad_fn=<NllLossBackward0>)
296000
29900 tensor(5.6953, device='cuda:0', grad_fn=<NllLossBackward0>)
297000
30000 tensor(5.3925, device='cuda:0', grad_fn=<NllLossBackward0>)
298000
30100 tensor(5.4241, device='cuda:0', grad_fn=<NllLossBackward0>)
299000
30200 tensor(5.4216, device='cuda:0', grad_fn=<NllLossBackward0>)
300000
30300 tensor(5.5074, device='cuda:0', grad_fn=<NllLossBackward0>)
301000
30400 tensor(5.3631, device='cuda:0', grad_fn=<NllLossBackward0>)
302000
30500 tensor(5.5690, device='cuda:0', grad_fn=<NllLossBackward0>)
30600 tensor(5.4734, device='cuda:0', grad_fn=<NllLossBackward0>)
303000
30700 tensor(5.5061, device='cuda:0', grad_fn=<NllLossBackward0>)
304000
30800 tensor(5.5709, device='cuda:0', grad_fn=<NllLossBackward0>)
305000
30900 tensor(5.5478, device='cuda:0', grad_fn=<NllLossBackward0>)
306000
31000 tensor(5.6687, device='cuda:0', grad_fn=<NllLossBackward0>)
307000
31100 tensor(5.2899, device='cuda:0', grad_fn=<NllLossBackward0>)
308000
31200 tensor(5.3663, device='cuda:0', grad_fn=<NllLossBackward0>)
309000
31300 tensor(5.6274, device='cuda:0', grad_fn=<NllLossBackward0>)
310000
31400 tensor(5.4358, device='cuda:0', grad_fn=<NllLossBackward0>)
311000
31500 tensor(5.5738, device='cuda:0', grad_fn=<NllLossBackward0>)
312000
31600 tensor(5.5612, device='cuda:0', grad_fn=<NllLossBackward0>)
313000
31700 tensor(5.5104, device='cuda:0', grad_fn=<NllLossBackward0>)
314000
31800 tensor(5.6343, device='cuda:0', grad_fn=<NllLossBackward0>)
315000
31900 tensor(5.2243, device='cuda:0', grad_fn=<NllLossBackward0>)
316000
32000 tensor(5.4320, device='cuda:0', grad_fn=<NllLossBackward0>)
317000
32100 tensor(5.3344, device='cuda:0', grad_fn=<NllLossBackward0>)
318000
32200 tensor(5.6543, device='cuda:0', grad_fn=<NllLossBackward0>)
319000
32300 tensor(5.6512, device='cuda:0', grad_fn=<NllLossBackward0>)
320000
32400 tensor(5.6237, device='cuda:0', grad_fn=<NllLossBackward0>)
321000
32500 tensor(5.4246, device='cuda:0', grad_fn=<NllLossBackward0>)
322000
32600 tensor(5.5469, device='cuda:0', grad_fn=<NllLossBackward0>)
323000
32700 tensor(5.5338, device='cuda:0', grad_fn=<NllLossBackward0>)
324000
32800 tensor(5.6954, device='cuda:0', grad_fn=<NllLossBackward0>)
325000
32900 tensor(5.5754, device='cuda:0', grad_fn=<NllLossBackward0>)
326000
33000 tensor(5.3334, device='cuda:0', grad_fn=<NllLossBackward0>)
327000
33100 tensor(5.5284, device='cuda:0', grad_fn=<NllLossBackward0>)
328000
33200 tensor(5.6350, device='cuda:0', grad_fn=<NllLossBackward0>)
329000
33300 tensor(5.4312, device='cuda:0', grad_fn=<NllLossBackward0>)
330000
33400 tensor(5.6854, device='cuda:0', grad_fn=<NllLossBackward0>)
331000
33500 tensor(5.4921, device='cuda:0', grad_fn=<NllLossBackward0>)
332000
33600 tensor(5.4345, device='cuda:0', grad_fn=<NllLossBackward0>)
333000
33700 tensor(5.4950, device='cuda:0', grad_fn=<NllLossBackward0>)
334000
33800 tensor(5.5757, device='cuda:0', grad_fn=<NllLossBackward0>)
335000
33900 tensor(5.3466, device='cuda:0', grad_fn=<NllLossBackward0>)
336000
34000 tensor(5.5373, device='cuda:0', grad_fn=<NllLossBackward0>)
337000
34100 tensor(5.5144, device='cuda:0', grad_fn=<NllLossBackward0>)
338000
34200 tensor(5.5543, device='cuda:0', grad_fn=<NllLossBackward0>)
339000
34300 tensor(5.3564, device='cuda:0', grad_fn=<NllLossBackward0>)
340000
34400 tensor(5.8091, device='cuda:0', grad_fn=<NllLossBackward0>)
341000
34500 tensor(5.6699, device='cuda:0', grad_fn=<NllLossBackward0>)
342000
34600 tensor(5.5536, device='cuda:0', grad_fn=<NllLossBackward0>)
343000
34700 tensor(5.6261, device='cuda:0', grad_fn=<NllLossBackward0>)
344000
34800 tensor(5.6504, device='cuda:0', grad_fn=<NllLossBackward0>)
345000
34900 tensor(5.7067, device='cuda:0', grad_fn=<NllLossBackward0>)
346000
35000 tensor(5.7307, device='cuda:0', grad_fn=<NllLossBackward0>)
347000
35100 tensor(5.4831, device='cuda:0', grad_fn=<NllLossBackward0>)
348000
35200 tensor(5.4367, device='cuda:0', grad_fn=<NllLossBackward0>)
349000
35300 tensor(5.6503, device='cuda:0', grad_fn=<NllLossBackward0>)
350000
35400 tensor(5.2892, device='cuda:0', grad_fn=<NllLossBackward0>)
351000
35500 tensor(5.4198, device='cuda:0', grad_fn=<NllLossBackward0>)
352000
35600 tensor(5.4870, device='cuda:0', grad_fn=<NllLossBackward0>)
353000
35700 tensor(5.4489, device='cuda:0', grad_fn=<NllLossBackward0>)
354000
35800 tensor(5.5170, device='cuda:0', grad_fn=<NllLossBackward0>)
355000
35900 tensor(5.4699, device='cuda:0', grad_fn=<NllLossBackward0>)
356000
36000 tensor(5.2451, device='cuda:0', grad_fn=<NllLossBackward0>)
357000
36100 tensor(5.6311, device='cuda:0', grad_fn=<NllLossBackward0>)
358000
36200 tensor(5.5157, device='cuda:0', grad_fn=<NllLossBackward0>)
359000
36300 tensor(5.7751, device='cuda:0', grad_fn=<NllLossBackward0>)
360000
36400 tensor(5.4740, device='cuda:0', grad_fn=<NllLossBackward0>)
361000
36500 tensor(5.4746, device='cuda:0', grad_fn=<NllLossBackward0>)
362000
36600 tensor(5.5244, device='cuda:0', grad_fn=<NllLossBackward0>)
363000
36700 tensor(5.3037, device='cuda:0', grad_fn=<NllLossBackward0>)
364000
36800 tensor(5.4238, device='cuda:0', grad_fn=<NllLossBackward0>)
365000
36900 tensor(5.5203, device='cuda:0', grad_fn=<NllLossBackward0>)
366000
37000 tensor(5.4431, device='cuda:0', grad_fn=<NllLossBackward0>)
367000
37100 tensor(5.4286, device='cuda:0', grad_fn=<NllLossBackward0>)
368000
37200 tensor(5.5108, device='cuda:0', grad_fn=<NllLossBackward0>)
369000
37300 tensor(5.4229, device='cuda:0', grad_fn=<NllLossBackward0>)
370000
37400 tensor(5.8406, device='cuda:0', grad_fn=<NllLossBackward0>)
371000
37500 tensor(5.4602, device='cuda:0', grad_fn=<NllLossBackward0>)
372000
37600 tensor(5.4417, device='cuda:0', grad_fn=<NllLossBackward0>)
373000
37700 tensor(5.6200, device='cuda:0', grad_fn=<NllLossBackward0>)
374000
37800 tensor(5.4527, device='cuda:0', grad_fn=<NllLossBackward0>)
375000
37900 tensor(5.4631, device='cuda:0', grad_fn=<NllLossBackward0>)
376000
38000 tensor(5.5196, device='cuda:0', grad_fn=<NllLossBackward0>)
377000
38100 tensor(5.5436, device='cuda:0', grad_fn=<NllLossBackward0>)
378000
38200 tensor(5.5269, device='cuda:0', grad_fn=<NllLossBackward0>)
379000
38300 tensor(5.4716, device='cuda:0', grad_fn=<NllLossBackward0>)
380000
38400 tensor(5.5081, device='cuda:0', grad_fn=<NllLossBackward0>)
381000
38500 tensor(5.5249, device='cuda:0', grad_fn=<NllLossBackward0>)
382000
38600 tensor(5.5018, device='cuda:0', grad_fn=<NllLossBackward0>)
383000
38700 tensor(5.4845, device='cuda:0', grad_fn=<NllLossBackward0>)
384000
38800 tensor(5.5505, device='cuda:0', grad_fn=<NllLossBackward0>)
385000
38900 tensor(5.6658, device='cuda:0', grad_fn=<NllLossBackward0>)
386000
39000 tensor(5.3333, device='cuda:0', grad_fn=<NllLossBackward0>)
387000
39100 tensor(5.5598, device='cuda:0', grad_fn=<NllLossBackward0>)
388000
39200 tensor(5.6624, device='cuda:0', grad_fn=<NllLossBackward0>)
389000
39300 tensor(5.4714, device='cuda:0', grad_fn=<NllLossBackward0>)
390000
39400 tensor(5.5470, device='cuda:0', grad_fn=<NllLossBackward0>)
391000
39500 tensor(5.6905, device='cuda:0', grad_fn=<NllLossBackward0>)
392000
39600 tensor(5.3592, device='cuda:0', grad_fn=<NllLossBackward0>)
393000
39700 tensor(5.3170, device='cuda:0', grad_fn=<NllLossBackward0>)
394000
39800 tensor(5.4491, device='cuda:0', grad_fn=<NllLossBackward0>)
395000
39900 tensor(5.2872, device='cuda:0', grad_fn=<NllLossBackward0>)
396000
40000 tensor(5.3865, device='cuda:0', grad_fn=<NllLossBackward0>)
397000
40100 tensor(5.4536, device='cuda:0', grad_fn=<NllLossBackward0>)
398000
40200 tensor(5.4382, device='cuda:0', grad_fn=<NllLossBackward0>)
399000
40300 tensor(5.4819, device='cuda:0', grad_fn=<NllLossBackward0>)
40400 tensor(5.5250, device='cuda:0', grad_fn=<NllLossBackward0>)
400000
40500 tensor(5.4396, device='cuda:0', grad_fn=<NllLossBackward0>)
401000
40600 tensor(5.5062, device='cuda:0', grad_fn=<NllLossBackward0>)
402000
40700 tensor(5.5362, device='cuda:0', grad_fn=<NllLossBackward0>)
403000
40800 tensor(5.5015, device='cuda:0', grad_fn=<NllLossBackward0>)
404000
40900 tensor(5.4610, device='cuda:0', grad_fn=<NllLossBackward0>)
405000
41000 tensor(5.5083, device='cuda:0', grad_fn=<NllLossBackward0>)
406000
41100 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>)
407000
41200 tensor(5.3340, device='cuda:0', grad_fn=<NllLossBackward0>)
408000
41300 tensor(5.4608, device='cuda:0', grad_fn=<NllLossBackward0>)
409000
41400 tensor(5.3758, device='cuda:0', grad_fn=<NllLossBackward0>)
410000
41500 tensor(5.5160, device='cuda:0', grad_fn=<NllLossBackward0>)
411000
41600 tensor(5.4290, device='cuda:0', grad_fn=<NllLossBackward0>)
412000
41700 tensor(5.4426, device='cuda:0', grad_fn=<NllLossBackward0>)
413000
41800 tensor(5.4764, device='cuda:0', grad_fn=<NllLossBackward0>)
414000
41900 tensor(5.4730, device='cuda:0', grad_fn=<NllLossBackward0>)
415000
42000 tensor(5.6150, device='cuda:0', grad_fn=<NllLossBackward0>)
416000
42100 tensor(5.3622, device='cuda:0', grad_fn=<NllLossBackward0>)
417000
42200 tensor(5.4380, device='cuda:0', grad_fn=<NllLossBackward0>)
418000
42300 tensor(5.5031, device='cuda:0', grad_fn=<NllLossBackward0>)
419000
42400 tensor(5.3124, device='cuda:0', grad_fn=<NllLossBackward0>)
420000
42500 tensor(5.4812, device='cuda:0', grad_fn=<NllLossBackward0>)
421000
42600 tensor(5.2723, device='cuda:0', grad_fn=<NllLossBackward0>)
422000
42700 tensor(5.5998, device='cuda:0', grad_fn=<NllLossBackward0>)
423000
42800 tensor(5.5254, device='cuda:0', grad_fn=<NllLossBackward0>)
424000
42900 tensor(5.3716, device='cuda:0', grad_fn=<NllLossBackward0>)
425000
43000 tensor(5.5020, device='cuda:0', grad_fn=<NllLossBackward0>)
426000
43100 tensor(5.5091, device='cuda:0', grad_fn=<NllLossBackward0>)
427000
43200 tensor(5.3182, device='cuda:0', grad_fn=<NllLossBackward0>)
428000
43300 tensor(5.4001, device='cuda:0', grad_fn=<NllLossBackward0>)
429000
43400 tensor(5.5150, device='cuda:0', grad_fn=<NllLossBackward0>)
430000
43500 tensor(5.2440, device='cuda:0', grad_fn=<NllLossBackward0>)
431000
43600 tensor(5.4439, device='cuda:0', grad_fn=<NllLossBackward0>)
432000
epoch: = 2
0 tensor(5.3953, device='cuda:0', grad_fn=<NllLossBackward0>)
1000
100 tensor(5.4847, device='cuda:0', grad_fn=<NllLossBackward0>)
200 tensor(5.3626, device='cuda:0', grad_fn=<NllLossBackward0>)
2000
300 tensor(5.4127, device='cuda:0', grad_fn=<NllLossBackward0>)
3000
400 tensor(5.3734, device='cuda:0', grad_fn=<NllLossBackward0>)
4000
500 tensor(5.5564, device='cuda:0', grad_fn=<NllLossBackward0>)
5000
600 tensor(5.3391, device='cuda:0', grad_fn=<NllLossBackward0>)
6000
700 tensor(5.6198, device='cuda:0', grad_fn=<NllLossBackward0>)
7000
800 tensor(5.2255, device='cuda:0', grad_fn=<NllLossBackward0>)
8000
900 tensor(5.5161, device='cuda:0', grad_fn=<NllLossBackward0>)
9000
1000 tensor(5.3517, device='cuda:0', grad_fn=<NllLossBackward0>)
10000
1100 tensor(5.5420, device='cuda:0', grad_fn=<NllLossBackward0>)
11000
1200 tensor(5.6031, device='cuda:0', grad_fn=<NllLossBackward0>)
12000
1300 tensor(5.5343, device='cuda:0', grad_fn=<NllLossBackward0>)
13000
1400 tensor(5.5547, device='cuda:0', grad_fn=<NllLossBackward0>)
14000
1500 tensor(5.6080, device='cuda:0', grad_fn=<NllLossBackward0>)
15000
1600 tensor(5.2940, device='cuda:0', grad_fn=<NllLossBackward0>)
16000
1700 tensor(5.3671, device='cuda:0', grad_fn=<NllLossBackward0>)
17000
1800 tensor(5.3777, device='cuda:0', grad_fn=<NllLossBackward0>)
18000
1900 tensor(5.3593, device='cuda:0', grad_fn=<NllLossBackward0>)
19000
2000 tensor(5.4348, device='cuda:0', grad_fn=<NllLossBackward0>)
20000
2100 tensor(5.5513, device='cuda:0', grad_fn=<NllLossBackward0>)
21000
2200 tensor(5.3939, device='cuda:0', grad_fn=<NllLossBackward0>)
22000
2300 tensor(5.4063, device='cuda:0', grad_fn=<NllLossBackward0>)
23000
2400 tensor(5.4092, device='cuda:0', grad_fn=<NllLossBackward0>)
24000
2500 tensor(5.4460, device='cuda:0', grad_fn=<NllLossBackward0>)
25000
2600 tensor(5.4738, device='cuda:0', grad_fn=<NllLossBackward0>)
26000
2700 tensor(5.4848, device='cuda:0', grad_fn=<NllLossBackward0>)
27000
2800 tensor(5.5244, device='cuda:0', grad_fn=<NllLossBackward0>)
28000
2900 tensor(5.6711, device='cuda:0', grad_fn=<NllLossBackward0>)
29000
3000 tensor(5.4024, device='cuda:0', grad_fn=<NllLossBackward0>)
30000
3100 tensor(5.4842, device='cuda:0', grad_fn=<NllLossBackward0>)
31000
3200 tensor(5.4863, device='cuda:0', grad_fn=<NllLossBackward0>)
32000
3300 tensor(5.4114, device='cuda:0', grad_fn=<NllLossBackward0>)
33000
3400 tensor(5.3231, device='cuda:0', grad_fn=<NllLossBackward0>)
34000
3500 tensor(5.4598, device='cuda:0', grad_fn=<NllLossBackward0>)
35000
3600 tensor(5.4579, device='cuda:0', grad_fn=<NllLossBackward0>)
36000
3700 tensor(5.3890, device='cuda:0', grad_fn=<NllLossBackward0>)
37000
3800 tensor(5.4162, device='cuda:0', grad_fn=<NllLossBackward0>)
38000
3900 tensor(5.2854, device='cuda:0', grad_fn=<NllLossBackward0>)
39000
4000 tensor(5.3370, device='cuda:0', grad_fn=<NllLossBackward0>)
40000
4100 tensor(5.5078, device='cuda:0', grad_fn=<NllLossBackward0>)
41000
4200 tensor(5.5341, device='cuda:0', grad_fn=<NllLossBackward0>)
42000
4300 tensor(5.4704, device='cuda:0', grad_fn=<NllLossBackward0>)
43000
4400 tensor(5.4990, device='cuda:0', grad_fn=<NllLossBackward0>)
44000
4500 tensor(5.3300, device='cuda:0', grad_fn=<NllLossBackward0>)
45000
4600 tensor(5.6674, device='cuda:0', grad_fn=<NllLossBackward0>)
46000
4700 tensor(5.3622, device='cuda:0', grad_fn=<NllLossBackward0>)
47000
4800 tensor(5.4762, device='cuda:0', grad_fn=<NllLossBackward0>)
48000
4900 tensor(5.5403, device='cuda:0', grad_fn=<NllLossBackward0>)
49000
5000 tensor(5.5359, device='cuda:0', grad_fn=<NllLossBackward0>)
50000
5100 tensor(5.6058, device='cuda:0', grad_fn=<NllLossBackward0>)
51000
5200 tensor(5.6209, device='cuda:0', grad_fn=<NllLossBackward0>)
52000
5300 tensor(5.6273, device='cuda:0', grad_fn=<NllLossBackward0>)
53000
5400 tensor(5.4695, device='cuda:0', grad_fn=<NllLossBackward0>)
54000
5500 tensor(5.5771, device='cuda:0', grad_fn=<NllLossBackward0>)
55000
5600 tensor(5.3552, device='cuda:0', grad_fn=<NllLossBackward0>)
56000
5700 tensor(5.5957, device='cuda:0', grad_fn=<NllLossBackward0>)
57000
5800 tensor(5.5952, device='cuda:0', grad_fn=<NllLossBackward0>)
58000
5900 tensor(5.5643, device='cuda:0', grad_fn=<NllLossBackward0>)
59000
6000 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>)
60000
6100 tensor(5.4620, device='cuda:0', grad_fn=<NllLossBackward0>)
61000
6200 tensor(5.6256, device='cuda:0', grad_fn=<NllLossBackward0>)
62000
6300 tensor(5.4832, device='cuda:0', grad_fn=<NllLossBackward0>)
63000
6400 tensor(5.4063, device='cuda:0', grad_fn=<NllLossBackward0>)
64000
6500 tensor(5.2587, device='cuda:0', grad_fn=<NllLossBackward0>)
65000
6600 tensor(5.4320, device='cuda:0', grad_fn=<NllLossBackward0>)
66000
6700 tensor(5.3770, device='cuda:0', grad_fn=<NllLossBackward0>)
67000
6800 tensor(5.6077, device='cuda:0', grad_fn=<NllLossBackward0>)
68000
6900 tensor(5.5788, device='cuda:0', grad_fn=<NllLossBackward0>)
69000
7000 tensor(5.4929, device='cuda:0', grad_fn=<NllLossBackward0>)
70000
7100 tensor(5.2828, device='cuda:0', grad_fn=<NllLossBackward0>)
71000
7200 tensor(5.3992, device='cuda:0', grad_fn=<NllLossBackward0>)
72000
7300 tensor(5.6273, device='cuda:0', grad_fn=<NllLossBackward0>)
73000
7400 tensor(5.4385, device='cuda:0', grad_fn=<NllLossBackward0>)
74000
7500 tensor(5.3176, device='cuda:0', grad_fn=<NllLossBackward0>)
75000
7600 tensor(5.3834, device='cuda:0', grad_fn=<NllLossBackward0>)
76000
7700 tensor(5.4532, device='cuda:0', grad_fn=<NllLossBackward0>)
77000
7800 tensor(5.5669, device='cuda:0', grad_fn=<NllLossBackward0>)
78000
7900 tensor(5.2508, device='cuda:0', grad_fn=<NllLossBackward0>)
79000
8000 tensor(5.3027, device='cuda:0', grad_fn=<NllLossBackward0>)
80000
8100 tensor(5.4813, device='cuda:0', grad_fn=<NllLossBackward0>)
81000
8200 tensor(5.4822, device='cuda:0', grad_fn=<NllLossBackward0>)
82000
8300 tensor(5.4510, device='cuda:0', grad_fn=<NllLossBackward0>)
83000
8400 tensor(5.5712, device='cuda:0', grad_fn=<NllLossBackward0>)
84000
8500 tensor(5.5634, device='cuda:0', grad_fn=<NllLossBackward0>)
85000
8600 tensor(5.5616, device='cuda:0', grad_fn=<NllLossBackward0>)
86000
8700 tensor(5.6568, device='cuda:0', grad_fn=<NllLossBackward0>)
87000
8800 tensor(5.5397, device='cuda:0', grad_fn=<NllLossBackward0>)
88000
8900 tensor(5.3852, device='cuda:0', grad_fn=<NllLossBackward0>)
89000
9000 tensor(5.5022, device='cuda:0', grad_fn=<NllLossBackward0>)
90000
9100 tensor(5.5088, device='cuda:0', grad_fn=<NllLossBackward0>)
91000
9200 tensor(5.4214, device='cuda:0', grad_fn=<NllLossBackward0>)
92000
9300 tensor(5.4641, device='cuda:0', grad_fn=<NllLossBackward0>)
93000
9400 tensor(5.3085, device='cuda:0', grad_fn=<NllLossBackward0>)
94000
9500 tensor(5.3852, device='cuda:0', grad_fn=<NllLossBackward0>)
95000
9600 tensor(5.5097, device='cuda:0', grad_fn=<NllLossBackward0>)
96000
9700 tensor(5.4373, device='cuda:0', grad_fn=<NllLossBackward0>)
97000
9800 tensor(5.4786, device='cuda:0', grad_fn=<NllLossBackward0>)
9900 tensor(5.3198, device='cuda:0', grad_fn=<NllLossBackward0>)
98000
10000 tensor(5.5310, device='cuda:0', grad_fn=<NllLossBackward0>)
99000
10100 tensor(5.4341, device='cuda:0', grad_fn=<NllLossBackward0>)
100000
10200 tensor(5.3571, device='cuda:0', grad_fn=<NllLossBackward0>)
101000
10300 tensor(5.4712, device='cuda:0', grad_fn=<NllLossBackward0>)
102000
10400 tensor(5.4810, device='cuda:0', grad_fn=<NllLossBackward0>)
103000
10500 tensor(5.5463, device='cuda:0', grad_fn=<NllLossBackward0>)
104000
10600 tensor(5.6233, device='cuda:0', grad_fn=<NllLossBackward0>)
105000
10700 tensor(5.4678, device='cuda:0', grad_fn=<NllLossBackward0>)
106000
10800 tensor(5.5040, device='cuda:0', grad_fn=<NllLossBackward0>)
107000
10900 tensor(5.3963, device='cuda:0', grad_fn=<NllLossBackward0>)
108000
11000 tensor(5.6295, device='cuda:0', grad_fn=<NllLossBackward0>)
109000
11100 tensor(5.2378, device='cuda:0', grad_fn=<NllLossBackward0>)
110000
11200 tensor(5.4184, device='cuda:0', grad_fn=<NllLossBackward0>)
111000
11300 tensor(5.4404, device='cuda:0', grad_fn=<NllLossBackward0>)
112000
11400 tensor(5.3875, device='cuda:0', grad_fn=<NllLossBackward0>)
113000
11500 tensor(5.4523, device='cuda:0', grad_fn=<NllLossBackward0>)
114000
11600 tensor(5.4418, device='cuda:0', grad_fn=<NllLossBackward0>)
115000
11700 tensor(5.3604, device='cuda:0', grad_fn=<NllLossBackward0>)
116000
11800 tensor(5.5647, device='cuda:0', grad_fn=<NllLossBackward0>)
117000
11900 tensor(5.3936, device='cuda:0', grad_fn=<NllLossBackward0>)
118000
12000 tensor(5.4823, device='cuda:0', grad_fn=<NllLossBackward0>)
119000
12100 tensor(5.5069, device='cuda:0', grad_fn=<NllLossBackward0>)
120000
12200 tensor(5.4983, device='cuda:0', grad_fn=<NllLossBackward0>)
121000
12300 tensor(5.6030, device='cuda:0', grad_fn=<NllLossBackward0>)
122000
12400 tensor(5.4763, device='cuda:0', grad_fn=<NllLossBackward0>)
123000
12500 tensor(5.3718, device='cuda:0', grad_fn=<NllLossBackward0>)
124000
12600 tensor(5.4416, device='cuda:0', grad_fn=<NllLossBackward0>)
125000
12700 tensor(5.3554, device='cuda:0', grad_fn=<NllLossBackward0>)
126000
12800 tensor(5.5392, device='cuda:0', grad_fn=<NllLossBackward0>)
127000
12900 tensor(5.4164, device='cuda:0', grad_fn=<NllLossBackward0>)
128000
13000 tensor(5.2286, device='cuda:0', grad_fn=<NllLossBackward0>)
129000
13100 tensor(5.4288, device='cuda:0', grad_fn=<NllLossBackward0>)
130000
13200 tensor(5.4770, device='cuda:0', grad_fn=<NllLossBackward0>)
131000
13300 tensor(5.3352, device='cuda:0', grad_fn=<NllLossBackward0>)
132000
13400 tensor(5.4349, device='cuda:0', grad_fn=<NllLossBackward0>)
133000
13500 tensor(5.3860, device='cuda:0', grad_fn=<NllLossBackward0>)
134000
13600 tensor(5.4648, device='cuda:0', grad_fn=<NllLossBackward0>)
135000
13700 tensor(5.4444, device='cuda:0', grad_fn=<NllLossBackward0>)
136000
13800 tensor(5.4320, device='cuda:0', grad_fn=<NllLossBackward0>)
137000
13900 tensor(5.2935, device='cuda:0', grad_fn=<NllLossBackward0>)
138000
14000 tensor(5.5387, device='cuda:0', grad_fn=<NllLossBackward0>)
139000
14100 tensor(5.2424, device='cuda:0', grad_fn=<NllLossBackward0>)
140000
14200 tensor(5.5177, device='cuda:0', grad_fn=<NllLossBackward0>)
141000
14300 tensor(5.4831, device='cuda:0', grad_fn=<NllLossBackward0>)
142000
14400 tensor(5.3877, device='cuda:0', grad_fn=<NllLossBackward0>)
143000
14500 tensor(5.4919, device='cuda:0', grad_fn=<NllLossBackward0>)
144000
14600 tensor(5.5253, device='cuda:0', grad_fn=<NllLossBackward0>)
145000
14700 tensor(5.3948, device='cuda:0', grad_fn=<NllLossBackward0>)
146000
14800 tensor(5.8442, device='cuda:0', grad_fn=<NllLossBackward0>)
147000
14900 tensor(5.4967, device='cuda:0', grad_fn=<NllLossBackward0>)
148000
15000 tensor(5.3788, device='cuda:0', grad_fn=<NllLossBackward0>)
149000
15100 tensor(5.5832, device='cuda:0', grad_fn=<NllLossBackward0>)
150000
15200 tensor(5.4482, device='cuda:0', grad_fn=<NllLossBackward0>)
151000
15300 tensor(5.4260, device='cuda:0', grad_fn=<NllLossBackward0>)
152000
15400 tensor(5.3273, device='cuda:0', grad_fn=<NllLossBackward0>)
153000
15500 tensor(5.4840, device='cuda:0', grad_fn=<NllLossBackward0>)
154000
15600 tensor(5.4851, device='cuda:0', grad_fn=<NllLossBackward0>)
155000
15700 tensor(5.3871, device='cuda:0', grad_fn=<NllLossBackward0>)
156000
15800 tensor(5.2933, device='cuda:0', grad_fn=<NllLossBackward0>)
157000
15900 tensor(5.4374, device='cuda:0', grad_fn=<NllLossBackward0>)
158000
16000 tensor(5.2555, device='cuda:0', grad_fn=<NllLossBackward0>)
159000
16100 tensor(5.3127, device='cuda:0', grad_fn=<NllLossBackward0>)
160000
16200 tensor(5.6423, device='cuda:0', grad_fn=<NllLossBackward0>)
161000
16300 tensor(5.4702, device='cuda:0', grad_fn=<NllLossBackward0>)
162000
16400 tensor(5.4419, device='cuda:0', grad_fn=<NllLossBackward0>)
163000
16500 tensor(5.5640, device='cuda:0', grad_fn=<NllLossBackward0>)
164000
16600 tensor(5.4099, device='cuda:0', grad_fn=<NllLossBackward0>)
165000
16700 tensor(5.3822, device='cuda:0', grad_fn=<NllLossBackward0>)
166000
16800 tensor(5.4643, device='cuda:0', grad_fn=<NllLossBackward0>)
167000
16900 tensor(5.2234, device='cuda:0', grad_fn=<NllLossBackward0>)
168000
17000 tensor(5.5021, device='cuda:0', grad_fn=<NllLossBackward0>)
169000
17100 tensor(5.3524, device='cuda:0', grad_fn=<NllLossBackward0>)
170000
17200 tensor(5.4725, device='cuda:0', grad_fn=<NllLossBackward0>)
171000
17300 tensor(5.5034, device='cuda:0', grad_fn=<NllLossBackward0>)
172000
17400 tensor(5.2911, device='cuda:0', grad_fn=<NllLossBackward0>)
173000
17500 tensor(5.3147, device='cuda:0', grad_fn=<NllLossBackward0>)
174000
17600 tensor(5.3426, device='cuda:0', grad_fn=<NllLossBackward0>)
175000
17700 tensor(5.3414, device='cuda:0', grad_fn=<NllLossBackward0>)
176000
17800 tensor(5.3991, device='cuda:0', grad_fn=<NllLossBackward0>)
177000
17900 tensor(5.2936, device='cuda:0', grad_fn=<NllLossBackward0>)
178000
18000 tensor(5.5238, device='cuda:0', grad_fn=<NllLossBackward0>)
179000
18100 tensor(5.4684, device='cuda:0', grad_fn=<NllLossBackward0>)
180000
18200 tensor(5.3916, device='cuda:0', grad_fn=<NllLossBackward0>)
181000
18300 tensor(5.3888, device='cuda:0', grad_fn=<NllLossBackward0>)
182000
18400 tensor(5.4299, device='cuda:0', grad_fn=<NllLossBackward0>)
183000
18500 tensor(5.4103, device='cuda:0', grad_fn=<NllLossBackward0>)
184000
18600 tensor(5.5980, device='cuda:0', grad_fn=<NllLossBackward0>)
185000
18700 tensor(5.4135, device='cuda:0', grad_fn=<NllLossBackward0>)
186000
18800 tensor(5.5855, device='cuda:0', grad_fn=<NllLossBackward0>)
187000
18900 tensor(5.4583, device='cuda:0', grad_fn=<NllLossBackward0>)
188000
19000 tensor(5.4854, device='cuda:0', grad_fn=<NllLossBackward0>)
189000
19100 tensor(5.5879, device='cuda:0', grad_fn=<NllLossBackward0>)
190000
19200 tensor(5.4675, device='cuda:0', grad_fn=<NllLossBackward0>)
191000
19300 tensor(5.5741, device='cuda:0', grad_fn=<NllLossBackward0>)
192000
19400 tensor(5.3977, device='cuda:0', grad_fn=<NllLossBackward0>)
193000
19500 tensor(5.4042, device='cuda:0', grad_fn=<NllLossBackward0>)
194000
19600 tensor(5.4364, device='cuda:0', grad_fn=<NllLossBackward0>)
195000
19700 tensor(5.4868, device='cuda:0', grad_fn=<NllLossBackward0>)
196000
19800 tensor(5.3476, device='cuda:0', grad_fn=<NllLossBackward0>)
197000
19900 tensor(5.3553, device='cuda:0', grad_fn=<NllLossBackward0>)
198000
20000 tensor(5.3707, device='cuda:0', grad_fn=<NllLossBackward0>)
199000
20100 tensor(5.3226, device='cuda:0', grad_fn=<NllLossBackward0>)
200000
20200 tensor(5.2488, device='cuda:0', grad_fn=<NllLossBackward0>)
201000
20300 tensor(5.3648, device='cuda:0', grad_fn=<NllLossBackward0>)
20400 tensor(5.4156, device='cuda:0', grad_fn=<NllLossBackward0>)
202000
20500 tensor(5.4102, device='cuda:0', grad_fn=<NllLossBackward0>)
203000
20600 tensor(5.6109, device='cuda:0', grad_fn=<NllLossBackward0>)
204000
20700 tensor(5.4335, device='cuda:0', grad_fn=<NllLossBackward0>)
205000
20800 tensor(5.2795, device='cuda:0', grad_fn=<NllLossBackward0>)
206000
20900 tensor(5.5609, device='cuda:0', grad_fn=<NllLossBackward0>)
207000
21000 tensor(5.3918, device='cuda:0', grad_fn=<NllLossBackward0>)
208000
21100 tensor(5.3831, device='cuda:0', grad_fn=<NllLossBackward0>)
209000
21200 tensor(5.2790, device='cuda:0', grad_fn=<NllLossBackward0>)
210000
21300 tensor(5.4710, device='cuda:0', grad_fn=<NllLossBackward0>)
211000
21400 tensor(5.5050, device='cuda:0', grad_fn=<NllLossBackward0>)
212000
21500 tensor(5.2692, device='cuda:0', grad_fn=<NllLossBackward0>)
213000
21600 tensor(5.2668, device='cuda:0', grad_fn=<NllLossBackward0>)
214000
21700 tensor(5.2633, device='cuda:0', grad_fn=<NllLossBackward0>)
215000
21800 tensor(5.4067, device='cuda:0', grad_fn=<NllLossBackward0>)
216000
21900 tensor(5.3829, device='cuda:0', grad_fn=<NllLossBackward0>)
217000
22000 tensor(5.3773, device='cuda:0', grad_fn=<NllLossBackward0>)
218000
22100 tensor(5.2472, device='cuda:0', grad_fn=<NllLossBackward0>)
219000
22200 tensor(5.1171, device='cuda:0', grad_fn=<NllLossBackward0>)
220000
22300 tensor(5.5545, device='cuda:0', grad_fn=<NllLossBackward0>)
221000
22400 tensor(5.2499, device='cuda:0', grad_fn=<NllLossBackward0>)
222000
22500 tensor(5.2943, device='cuda:0', grad_fn=<NllLossBackward0>)
223000
22600 tensor(5.4748, device='cuda:0', grad_fn=<NllLossBackward0>)
224000
22700 tensor(5.2436, device='cuda:0', grad_fn=<NllLossBackward0>)
225000
22800 tensor(5.5053, device='cuda:0', grad_fn=<NllLossBackward0>)
226000
22900 tensor(5.5519, device='cuda:0', grad_fn=<NllLossBackward0>)
227000
23000 tensor(5.4541, device='cuda:0', grad_fn=<NllLossBackward0>)
228000
23100 tensor(5.4279, device='cuda:0', grad_fn=<NllLossBackward0>)
229000
23200 tensor(5.4286, device='cuda:0', grad_fn=<NllLossBackward0>)
230000
23300 tensor(5.5179, device='cuda:0', grad_fn=<NllLossBackward0>)
231000
23400 tensor(5.5355, device='cuda:0', grad_fn=<NllLossBackward0>)
232000
23500 tensor(5.3505, device='cuda:0', grad_fn=<NllLossBackward0>)
233000
23600 tensor(5.3313, device='cuda:0', grad_fn=<NllLossBackward0>)
234000
23700 tensor(5.3509, device='cuda:0', grad_fn=<NllLossBackward0>)
235000
23800 tensor(5.2170, device='cuda:0', grad_fn=<NllLossBackward0>)
236000
23900 tensor(5.3101, device='cuda:0', grad_fn=<NllLossBackward0>)
237000
24000 tensor(5.2962, device='cuda:0', grad_fn=<NllLossBackward0>)
238000
24100 tensor(5.3882, device='cuda:0', grad_fn=<NllLossBackward0>)
239000
24200 tensor(5.5633, device='cuda:0', grad_fn=<NllLossBackward0>)
240000
24300 tensor(5.4595, device='cuda:0', grad_fn=<NllLossBackward0>)
241000
24400 tensor(5.5932, device='cuda:0', grad_fn=<NllLossBackward0>)
242000
24500 tensor(5.3717, device='cuda:0', grad_fn=<NllLossBackward0>)
243000
24600 tensor(5.4943, device='cuda:0', grad_fn=<NllLossBackward0>)
244000
24700 tensor(5.3985, device='cuda:0', grad_fn=<NllLossBackward0>)
245000
24800 tensor(5.4347, device='cuda:0', grad_fn=<NllLossBackward0>)
246000
24900 tensor(5.5008, device='cuda:0', grad_fn=<NllLossBackward0>)
247000
25000 tensor(5.5100, device='cuda:0', grad_fn=<NllLossBackward0>)
248000
25100 tensor(5.4427, device='cuda:0', grad_fn=<NllLossBackward0>)
249000
25200 tensor(5.4508, device='cuda:0', grad_fn=<NllLossBackward0>)
250000
25300 tensor(5.4724, device='cuda:0', grad_fn=<NllLossBackward0>)
251000
25400 tensor(5.4525, device='cuda:0', grad_fn=<NllLossBackward0>)
252000
25500 tensor(5.3620, device='cuda:0', grad_fn=<NllLossBackward0>)
253000
25600 tensor(5.6446, device='cuda:0', grad_fn=<NllLossBackward0>)
254000
25700 tensor(5.3966, device='cuda:0', grad_fn=<NllLossBackward0>)
255000
25800 tensor(5.4889, device='cuda:0', grad_fn=<NllLossBackward0>)
256000
25900 tensor(5.4251, device='cuda:0', grad_fn=<NllLossBackward0>)
257000
26000 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>)
258000
26100 tensor(5.3395, device='cuda:0', grad_fn=<NllLossBackward0>)
259000
26200 tensor(5.2695, device='cuda:0', grad_fn=<NllLossBackward0>)
260000
26300 tensor(5.4767, device='cuda:0', grad_fn=<NllLossBackward0>)
261000
26400 tensor(5.5083, device='cuda:0', grad_fn=<NllLossBackward0>)
262000
26500 tensor(5.2347, device='cuda:0', grad_fn=<NllLossBackward0>)
263000
26600 tensor(5.5761, device='cuda:0', grad_fn=<NllLossBackward0>)
264000
26700 tensor(5.4402, device='cuda:0', grad_fn=<NllLossBackward0>)
265000
26800 tensor(5.6173, device='cuda:0', grad_fn=<NllLossBackward0>)
266000
26900 tensor(5.3775, device='cuda:0', grad_fn=<NllLossBackward0>)
267000
27000 tensor(5.2863, device='cuda:0', grad_fn=<NllLossBackward0>)
268000
27100 tensor(5.3007, device='cuda:0', grad_fn=<NllLossBackward0>)
269000
27200 tensor(5.3551, device='cuda:0', grad_fn=<NllLossBackward0>)
270000
27300 tensor(5.5439, device='cuda:0', grad_fn=<NllLossBackward0>)
271000
27400 tensor(5.4334, device='cuda:0', grad_fn=<NllLossBackward0>)
272000
27500 tensor(5.3266, device='cuda:0', grad_fn=<NllLossBackward0>)
273000
27600 tensor(5.6412, device='cuda:0', grad_fn=<NllLossBackward0>)
274000
27700 tensor(5.4420, device='cuda:0', grad_fn=<NllLossBackward0>)
275000
27800 tensor(5.4381, device='cuda:0', grad_fn=<NllLossBackward0>)
276000
27900 tensor(5.5550, device='cuda:0', grad_fn=<NllLossBackward0>)
277000
28000 tensor(5.4154, device='cuda:0', grad_fn=<NllLossBackward0>)
278000
28100 tensor(5.3823, device='cuda:0', grad_fn=<NllLossBackward0>)
279000
28200 tensor(5.5344, device='cuda:0', grad_fn=<NllLossBackward0>)
280000
28300 tensor(5.1615, device='cuda:0', grad_fn=<NllLossBackward0>)
281000
28400 tensor(5.6069, device='cuda:0', grad_fn=<NllLossBackward0>)
282000
28500 tensor(5.4426, device='cuda:0', grad_fn=<NllLossBackward0>)
283000
28600 tensor(5.3672, device='cuda:0', grad_fn=<NllLossBackward0>)
284000
28700 tensor(5.5133, device='cuda:0', grad_fn=<NllLossBackward0>)
285000
28800 tensor(5.5556, device='cuda:0', grad_fn=<NllLossBackward0>)
286000
28900 tensor(5.4294, device='cuda:0', grad_fn=<NllLossBackward0>)
287000
29000 tensor(5.3359, device='cuda:0', grad_fn=<NllLossBackward0>)
288000
29100 tensor(5.0951, device='cuda:0', grad_fn=<NllLossBackward0>)
289000
29200 tensor(5.2511, device='cuda:0', grad_fn=<NllLossBackward0>)
290000
29300 tensor(5.5364, device='cuda:0', grad_fn=<NllLossBackward0>)
291000
29400 tensor(5.6708, device='cuda:0', grad_fn=<NllLossBackward0>)
292000
29500 tensor(5.4371, device='cuda:0', grad_fn=<NllLossBackward0>)
293000
29600 tensor(5.2942, device='cuda:0', grad_fn=<NllLossBackward0>)
294000
29700 tensor(5.4637, device='cuda:0', grad_fn=<NllLossBackward0>)
295000
29800 tensor(5.2914, device='cuda:0', grad_fn=<NllLossBackward0>)
296000
29900 tensor(5.5562, device='cuda:0', grad_fn=<NllLossBackward0>)
297000
30000 tensor(5.2833, device='cuda:0', grad_fn=<NllLossBackward0>)
298000
30100 tensor(5.3481, device='cuda:0', grad_fn=<NllLossBackward0>)
299000
30200 tensor(5.3122, device='cuda:0', grad_fn=<NllLossBackward0>)
300000
30300 tensor(5.4103, device='cuda:0', grad_fn=<NllLossBackward0>)
301000
30400 tensor(5.2480, device='cuda:0', grad_fn=<NllLossBackward0>)
302000
30500 tensor(5.4258, device='cuda:0', grad_fn=<NllLossBackward0>)
30600 tensor(5.3835, device='cuda:0', grad_fn=<NllLossBackward0>)
303000
30700 tensor(5.4193, device='cuda:0', grad_fn=<NllLossBackward0>)
304000
30800 tensor(5.4438, device='cuda:0', grad_fn=<NllLossBackward0>)
305000
30900 tensor(5.4518, device='cuda:0', grad_fn=<NllLossBackward0>)
306000
31000 tensor(5.5607, device='cuda:0', grad_fn=<NllLossBackward0>)
307000
31100 tensor(5.2059, device='cuda:0', grad_fn=<NllLossBackward0>)
308000
31200 tensor(5.2571, device='cuda:0', grad_fn=<NllLossBackward0>)
309000
31300 tensor(5.5208, device='cuda:0', grad_fn=<NllLossBackward0>)
310000
31400 tensor(5.3061, device='cuda:0', grad_fn=<NllLossBackward0>)
311000
31500 tensor(5.4834, device='cuda:0', grad_fn=<NllLossBackward0>)
312000
31600 tensor(5.4653, device='cuda:0', grad_fn=<NllLossBackward0>)
313000
31700 tensor(5.4308, device='cuda:0', grad_fn=<NllLossBackward0>)
314000
31800 tensor(5.5400, device='cuda:0', grad_fn=<NllLossBackward0>)
315000
31900 tensor(5.1536, device='cuda:0', grad_fn=<NllLossBackward0>)
316000
32000 tensor(5.3460, device='cuda:0', grad_fn=<NllLossBackward0>)
317000
32100 tensor(5.2300, device='cuda:0', grad_fn=<NllLossBackward0>)
318000
32200 tensor(5.5511, device='cuda:0', grad_fn=<NllLossBackward0>)
319000
32300 tensor(5.5391, device='cuda:0', grad_fn=<NllLossBackward0>)
320000
32400 tensor(5.5157, device='cuda:0', grad_fn=<NllLossBackward0>)
321000
32500 tensor(5.3336, device='cuda:0', grad_fn=<NllLossBackward0>)
322000
32600 tensor(5.4475, device='cuda:0', grad_fn=<NllLossBackward0>)
323000
32700 tensor(5.3894, device='cuda:0', grad_fn=<NllLossBackward0>)
324000
32800 tensor(5.6022, device='cuda:0', grad_fn=<NllLossBackward0>)
325000
32900 tensor(5.4663, device='cuda:0', grad_fn=<NllLossBackward0>)
326000
33000 tensor(5.2387, device='cuda:0', grad_fn=<NllLossBackward0>)
327000
33100 tensor(5.4446, device='cuda:0', grad_fn=<NllLossBackward0>)
328000
33200 tensor(5.5450, device='cuda:0', grad_fn=<NllLossBackward0>)
329000
33300 tensor(5.3179, device='cuda:0', grad_fn=<NllLossBackward0>)
330000
33400 tensor(5.5905, device='cuda:0', grad_fn=<NllLossBackward0>)
331000
33500 tensor(5.4066, device='cuda:0', grad_fn=<NllLossBackward0>)
332000
33600 tensor(5.3542, device='cuda:0', grad_fn=<NllLossBackward0>)
333000
33700 tensor(5.4097, device='cuda:0', grad_fn=<NllLossBackward0>)
334000
33800 tensor(5.4912, device='cuda:0', grad_fn=<NllLossBackward0>)
335000
33900 tensor(5.2358, device='cuda:0', grad_fn=<NllLossBackward0>)
336000
34000 tensor(5.4470, device='cuda:0', grad_fn=<NllLossBackward0>)
337000
34100 tensor(5.4207, device='cuda:0', grad_fn=<NllLossBackward0>)
338000
34200 tensor(5.4651, device='cuda:0', grad_fn=<NllLossBackward0>)
339000
34300 tensor(5.2545, device='cuda:0', grad_fn=<NllLossBackward0>)
340000
34400 tensor(5.7106, device='cuda:0', grad_fn=<NllLossBackward0>)
341000
34500 tensor(5.5699, device='cuda:0', grad_fn=<NllLossBackward0>)
342000
34600 tensor(5.4638, device='cuda:0', grad_fn=<NllLossBackward0>)
343000
34700 tensor(5.5382, device='cuda:0', grad_fn=<NllLossBackward0>)
344000
34800 tensor(5.5603, device='cuda:0', grad_fn=<NllLossBackward0>)
345000
34900 tensor(5.6072, device='cuda:0', grad_fn=<NllLossBackward0>)
346000
35000 tensor(5.6037, device='cuda:0', grad_fn=<NllLossBackward0>)
347000
35100 tensor(5.4069, device='cuda:0', grad_fn=<NllLossBackward0>)
348000
35200 tensor(5.3398, device='cuda:0', grad_fn=<NllLossBackward0>)
349000
35300 tensor(5.5607, device='cuda:0', grad_fn=<NllLossBackward0>)
350000
35400 tensor(5.2068, device='cuda:0', grad_fn=<NllLossBackward0>)
351000
35500 tensor(5.3112, device='cuda:0', grad_fn=<NllLossBackward0>)
352000
35600 tensor(5.4126, device='cuda:0', grad_fn=<NllLossBackward0>)
353000
35700 tensor(5.3091, device='cuda:0', grad_fn=<NllLossBackward0>)
354000
35800 tensor(5.4252, device='cuda:0', grad_fn=<NllLossBackward0>)
355000
35900 tensor(5.3956, device='cuda:0', grad_fn=<NllLossBackward0>)
356000
36000 tensor(5.1705, device='cuda:0', grad_fn=<NllLossBackward0>)
357000
36100 tensor(5.5497, device='cuda:0', grad_fn=<NllLossBackward0>)
358000
36200 tensor(5.4066, device='cuda:0', grad_fn=<NllLossBackward0>)
359000
36300 tensor(5.6858, device='cuda:0', grad_fn=<NllLossBackward0>)
360000
36400 tensor(5.3812, device='cuda:0', grad_fn=<NllLossBackward0>)
361000
36500 tensor(5.3990, device='cuda:0', grad_fn=<NllLossBackward0>)
362000
36600 tensor(5.4302, device='cuda:0', grad_fn=<NllLossBackward0>)
363000
36700 tensor(5.2253, device='cuda:0', grad_fn=<NllLossBackward0>)
364000
36800 tensor(5.3347, device='cuda:0', grad_fn=<NllLossBackward0>)
365000
36900 tensor(5.4426, device='cuda:0', grad_fn=<NllLossBackward0>)
366000
37000 tensor(5.3419, device='cuda:0', grad_fn=<NllLossBackward0>)
367000
37100 tensor(5.3579, device='cuda:0', grad_fn=<NllLossBackward0>)
368000
37200 tensor(5.4332, device='cuda:0', grad_fn=<NllLossBackward0>)
369000
37300 tensor(5.3362, device='cuda:0', grad_fn=<NllLossBackward0>)
370000
37400 tensor(5.7100, device='cuda:0', grad_fn=<NllLossBackward0>)
371000
37500 tensor(5.3742, device='cuda:0', grad_fn=<NllLossBackward0>)
372000
37600 tensor(5.3615, device='cuda:0', grad_fn=<NllLossBackward0>)
373000
37700 tensor(5.5402, device='cuda:0', grad_fn=<NllLossBackward0>)
374000
37800 tensor(5.3734, device='cuda:0', grad_fn=<NllLossBackward0>)
375000
37900 tensor(5.3621, device='cuda:0', grad_fn=<NllLossBackward0>)
376000
38000 tensor(5.4380, device='cuda:0', grad_fn=<NllLossBackward0>)
377000
38100 tensor(5.4513, device='cuda:0', grad_fn=<NllLossBackward0>)
378000
38200 tensor(5.4554, device='cuda:0', grad_fn=<NllLossBackward0>)
379000
38300 tensor(5.3735, device='cuda:0', grad_fn=<NllLossBackward0>)
380000
38400 tensor(5.4297, device='cuda:0', grad_fn=<NllLossBackward0>)
381000
38500 tensor(5.4561, device='cuda:0', grad_fn=<NllLossBackward0>)
382000
38600 tensor(5.4118, device='cuda:0', grad_fn=<NllLossBackward0>)
383000
38700 tensor(5.3996, device='cuda:0', grad_fn=<NllLossBackward0>)
384000
38800 tensor(5.4825, device='cuda:0', grad_fn=<NllLossBackward0>)
385000
38900 tensor(5.5692, device='cuda:0', grad_fn=<NllLossBackward0>)
386000
39000 tensor(5.2573, device='cuda:0', grad_fn=<NllLossBackward0>)
387000
39100 tensor(5.4847, device='cuda:0', grad_fn=<NllLossBackward0>)
388000
39200 tensor(5.5802, device='cuda:0', grad_fn=<NllLossBackward0>)
389000
39300 tensor(5.3968, device='cuda:0', grad_fn=<NllLossBackward0>)
390000
39400 tensor(5.4666, device='cuda:0', grad_fn=<NllLossBackward0>)
391000
39500 tensor(5.5847, device='cuda:0', grad_fn=<NllLossBackward0>)
392000
39600 tensor(5.2648, device='cuda:0', grad_fn=<NllLossBackward0>)
393000
39700 tensor(5.2423, device='cuda:0', grad_fn=<NllLossBackward0>)
394000
39800 tensor(5.3731, device='cuda:0', grad_fn=<NllLossBackward0>)
395000
39900 tensor(5.2014, device='cuda:0', grad_fn=<NllLossBackward0>)
396000
40000 tensor(5.2903, device='cuda:0', grad_fn=<NllLossBackward0>)
397000
40100 tensor(5.3712, device='cuda:0', grad_fn=<NllLossBackward0>)
398000
40200 tensor(5.3557, device='cuda:0', grad_fn=<NllLossBackward0>)
399000
40300 tensor(5.4151, device='cuda:0', grad_fn=<NllLossBackward0>)
40400 tensor(5.4358, device='cuda:0', grad_fn=<NllLossBackward0>)
400000
40500 tensor(5.3498, device='cuda:0', grad_fn=<NllLossBackward0>)
401000
40600 tensor(5.4152, device='cuda:0', grad_fn=<NllLossBackward0>)
402000
40700 tensor(5.4551, device='cuda:0', grad_fn=<NllLossBackward0>)
403000
40800 tensor(5.4138, device='cuda:0', grad_fn=<NllLossBackward0>)
404000
40900 tensor(5.3628, device='cuda:0', grad_fn=<NllLossBackward0>)
405000
41000 tensor(5.4124, device='cuda:0', grad_fn=<NllLossBackward0>)
406000
41100 tensor(5.3750, device='cuda:0', grad_fn=<NllLossBackward0>)
407000
41200 tensor(5.2687, device='cuda:0', grad_fn=<NllLossBackward0>)
408000
41300 tensor(5.3987, device='cuda:0', grad_fn=<NllLossBackward0>)
409000
41400 tensor(5.2976, device='cuda:0', grad_fn=<NllLossBackward0>)
410000
41500 tensor(5.4418, device='cuda:0', grad_fn=<NllLossBackward0>)
411000
41600 tensor(5.3558, device='cuda:0', grad_fn=<NllLossBackward0>)
412000
41700 tensor(5.3767, device='cuda:0', grad_fn=<NllLossBackward0>)
413000
41800 tensor(5.3836, device='cuda:0', grad_fn=<NllLossBackward0>)
414000
41900 tensor(5.3904, device='cuda:0', grad_fn=<NllLossBackward0>)
415000
42000 tensor(5.5445, device='cuda:0', grad_fn=<NllLossBackward0>)
416000
42100 tensor(5.2890, device='cuda:0', grad_fn=<NllLossBackward0>)
417000
42200 tensor(5.3691, device='cuda:0', grad_fn=<NllLossBackward0>)
418000
42300 tensor(5.4364, device='cuda:0', grad_fn=<NllLossBackward0>)
419000
42400 tensor(5.2507, device='cuda:0', grad_fn=<NllLossBackward0>)
420000
42500 tensor(5.4215, device='cuda:0', grad_fn=<NllLossBackward0>)
421000
42600 tensor(5.2136, device='cuda:0', grad_fn=<NllLossBackward0>)
422000
42700 tensor(5.5296, device='cuda:0', grad_fn=<NllLossBackward0>)
423000
42800 tensor(5.4544, device='cuda:0', grad_fn=<NllLossBackward0>)
424000
42900 tensor(5.3009, device='cuda:0', grad_fn=<NllLossBackward0>)
425000
43000 tensor(5.4403, device='cuda:0', grad_fn=<NllLossBackward0>)
426000
43100 tensor(5.4384, device='cuda:0', grad_fn=<NllLossBackward0>)
427000
43200 tensor(5.2520, device='cuda:0', grad_fn=<NllLossBackward0>)
428000
43300 tensor(5.2945, device='cuda:0', grad_fn=<NllLossBackward0>)
429000
43400 tensor(5.4455, device='cuda:0', grad_fn=<NllLossBackward0>)
430000
43500 tensor(5.1633, device='cuda:0', grad_fn=<NllLossBackward0>)
431000
43600 tensor(5.3649, device='cuda:0', grad_fn=<NllLossBackward0>)
432000
epoch: = 3
0 tensor(5.3427, device='cuda:0', grad_fn=<NllLossBackward0>)
1000
100 tensor(5.4180, device='cuda:0', grad_fn=<NllLossBackward0>)
200 tensor(5.2939, device='cuda:0', grad_fn=<NllLossBackward0>)
2000
300 tensor(5.3083, device='cuda:0', grad_fn=<NllLossBackward0>)
3000
400 tensor(5.3086, device='cuda:0', grad_fn=<NllLossBackward0>)
4000
500 tensor(5.4733, device='cuda:0', grad_fn=<NllLossBackward0>)
5000
600 tensor(5.2627, device='cuda:0', grad_fn=<NllLossBackward0>)
6000
700 tensor(5.5664, device='cuda:0', grad_fn=<NllLossBackward0>)
7000
800 tensor(5.1641, device='cuda:0', grad_fn=<NllLossBackward0>)
8000
900 tensor(5.4272, device='cuda:0', grad_fn=<NllLossBackward0>)
9000
1000 tensor(5.2926, device='cuda:0', grad_fn=<NllLossBackward0>)
10000
1100 tensor(5.4848, device='cuda:0', grad_fn=<NllLossBackward0>)
11000
1200 tensor(5.5283, device='cuda:0', grad_fn=<NllLossBackward0>)
12000
1300 tensor(5.4635, device='cuda:0', grad_fn=<NllLossBackward0>)
13000
1400 tensor(5.4590, device='cuda:0', grad_fn=<NllLossBackward0>)
14000
1500 tensor(5.5386, device='cuda:0', grad_fn=<NllLossBackward0>)
15000
1600 tensor(5.2150, device='cuda:0', grad_fn=<NllLossBackward0>)
16000
1700 tensor(5.3116, device='cuda:0', grad_fn=<NllLossBackward0>)
17000
1800 tensor(5.3130, device='cuda:0', grad_fn=<NllLossBackward0>)
18000
1900 tensor(5.2889, device='cuda:0', grad_fn=<NllLossBackward0>)
19000
2000 tensor(5.3574, device='cuda:0', grad_fn=<NllLossBackward0>)
20000
2100 tensor(5.4860, device='cuda:0', grad_fn=<NllLossBackward0>)
21000
2200 tensor(5.3206, device='cuda:0', grad_fn=<NllLossBackward0>)
22000
2300 tensor(5.3447, device='cuda:0', grad_fn=<NllLossBackward0>)
23000
2400 tensor(5.3333, device='cuda:0', grad_fn=<NllLossBackward0>)
24000
2500 tensor(5.3822, device='cuda:0', grad_fn=<NllLossBackward0>)
25000
2600 tensor(5.4039, device='cuda:0', grad_fn=<NllLossBackward0>)
26000
2700 tensor(5.4280, device='cuda:0', grad_fn=<NllLossBackward0>)
27000
2800 tensor(5.4575, device='cuda:0', grad_fn=<NllLossBackward0>)
28000
2900 tensor(5.5878, device='cuda:0', grad_fn=<NllLossBackward0>)
29000
3000 tensor(5.3311, device='cuda:0', grad_fn=<NllLossBackward0>)
30000
3100 tensor(5.4103, device='cuda:0', grad_fn=<NllLossBackward0>)
31000
3200 tensor(5.4323, device='cuda:0', grad_fn=<NllLossBackward0>)
32000
3300 tensor(5.3521, device='cuda:0', grad_fn=<NllLossBackward0>)
33000
3400 tensor(5.2512, device='cuda:0', grad_fn=<NllLossBackward0>)
34000
3500 tensor(5.3813, device='cuda:0', grad_fn=<NllLossBackward0>)
35000
3600 tensor(5.4000, device='cuda:0', grad_fn=<NllLossBackward0>)
36000
3700 tensor(5.3312, device='cuda:0', grad_fn=<NllLossBackward0>)
37000
3800 tensor(5.3553, device='cuda:0', grad_fn=<NllLossBackward0>)
38000
3900 tensor(5.2275, device='cuda:0', grad_fn=<NllLossBackward0>)
39000
4000 tensor(5.2883, device='cuda:0', grad_fn=<NllLossBackward0>)
40000
4100 tensor(5.4294, device='cuda:0', grad_fn=<NllLossBackward0>)
41000
4200 tensor(5.4801, device='cuda:0', grad_fn=<NllLossBackward0>)
42000
4300 tensor(5.3863, device='cuda:0', grad_fn=<NllLossBackward0>)
43000
4400 tensor(5.4470, device='cuda:0', grad_fn=<NllLossBackward0>)
44000
4500 tensor(5.2610, device='cuda:0', grad_fn=<NllLossBackward0>)
45000
4600 tensor(5.5962, device='cuda:0', grad_fn=<NllLossBackward0>)
46000
4700 tensor(5.3029, device='cuda:0', grad_fn=<NllLossBackward0>)
47000
4800 tensor(5.4265, device='cuda:0', grad_fn=<NllLossBackward0>)
48000
4900 tensor(5.4823, device='cuda:0', grad_fn=<NllLossBackward0>)
49000
5000 tensor(5.4749, device='cuda:0', grad_fn=<NllLossBackward0>)
50000
5100 tensor(5.5356, device='cuda:0', grad_fn=<NllLossBackward0>)
51000
5200 tensor(5.5513, device='cuda:0', grad_fn=<NllLossBackward0>)
52000
5300 tensor(5.5476, device='cuda:0', grad_fn=<NllLossBackward0>)
53000
5400 tensor(5.4039, device='cuda:0', grad_fn=<NllLossBackward0>)
54000
5500 tensor(5.5156, device='cuda:0', grad_fn=<NllLossBackward0>)
55000
5600 tensor(5.2975, device='cuda:0', grad_fn=<NllLossBackward0>)
56000
5700 tensor(5.5492, device='cuda:0', grad_fn=<NllLossBackward0>)
57000
5800 tensor(5.5379, device='cuda:0', grad_fn=<NllLossBackward0>)
58000
5900 tensor(5.4874, device='cuda:0', grad_fn=<NllLossBackward0>)
59000
6000 tensor(5.3808, device='cuda:0', grad_fn=<NllLossBackward0>)
60000
6100 tensor(5.3932, device='cuda:0', grad_fn=<NllLossBackward0>)
61000
6200 tensor(5.5657, device='cuda:0', grad_fn=<NllLossBackward0>)
62000
6300 tensor(5.4233, device='cuda:0', grad_fn=<NllLossBackward0>)
63000
6400 tensor(5.3438, device='cuda:0', grad_fn=<NllLossBackward0>)
64000
6500 tensor(5.2002, device='cuda:0', grad_fn=<NllLossBackward0>)
65000
6600 tensor(5.3774, device='cuda:0', grad_fn=<NllLossBackward0>)
66000
6700 tensor(5.3193, device='cuda:0', grad_fn=<NllLossBackward0>)
67000
6800 tensor(5.5394, device='cuda:0', grad_fn=<NllLossBackward0>)
68000
6900 tensor(5.5196, device='cuda:0', grad_fn=<NllLossBackward0>)
69000
7000 tensor(5.4282, device='cuda:0', grad_fn=<NllLossBackward0>)
70000
7100 tensor(5.2296, device='cuda:0', grad_fn=<NllLossBackward0>)
71000
7200 tensor(5.3175, device='cuda:0', grad_fn=<NllLossBackward0>)
72000
7300 tensor(5.5642, device='cuda:0', grad_fn=<NllLossBackward0>)
73000
7400 tensor(5.3784, device='cuda:0', grad_fn=<NllLossBackward0>)
74000
7500 tensor(5.2475, device='cuda:0', grad_fn=<NllLossBackward0>)
75000
7600 tensor(5.3194, device='cuda:0', grad_fn=<NllLossBackward0>)
76000
7700 tensor(5.3934, device='cuda:0', grad_fn=<NllLossBackward0>)
77000
7800 tensor(5.5041, device='cuda:0', grad_fn=<NllLossBackward0>)
78000
7900 tensor(5.1814, device='cuda:0', grad_fn=<NllLossBackward0>)
79000
8000 tensor(5.2426, device='cuda:0', grad_fn=<NllLossBackward0>)
80000
8100 tensor(5.4104, device='cuda:0', grad_fn=<NllLossBackward0>)
81000
8200 tensor(5.4198, device='cuda:0', grad_fn=<NllLossBackward0>)
82000
8300 tensor(5.3854, device='cuda:0', grad_fn=<NllLossBackward0>)
83000
8400 tensor(5.5128, device='cuda:0', grad_fn=<NllLossBackward0>)
84000
8500 tensor(5.4898, device='cuda:0', grad_fn=<NllLossBackward0>)
85000
8600 tensor(5.4943, device='cuda:0', grad_fn=<NllLossBackward0>)
86000
8700 tensor(5.6012, device='cuda:0', grad_fn=<NllLossBackward0>)
87000
8800 tensor(5.4790, device='cuda:0', grad_fn=<NllLossBackward0>)
88000
8900 tensor(5.3312, device='cuda:0', grad_fn=<NllLossBackward0>)
89000
9000 tensor(5.4456, device='cuda:0', grad_fn=<NllLossBackward0>)
90000
9100 tensor(5.4537, device='cuda:0', grad_fn=<NllLossBackward0>)
91000
9200 tensor(5.3643, device='cuda:0', grad_fn=<NllLossBackward0>)
92000
9300 tensor(5.4085, device='cuda:0', grad_fn=<NllLossBackward0>)
93000
9400 tensor(5.2527, device='cuda:0', grad_fn=<NllLossBackward0>)
94000
9500 tensor(5.3289, device='cuda:0', grad_fn=<NllLossBackward0>)
95000
9600 tensor(5.4516, device='cuda:0', grad_fn=<NllLossBackward0>)
96000
9700 tensor(5.3881, device='cuda:0', grad_fn=<NllLossBackward0>)
97000
9800 tensor(5.4321, device='cuda:0', grad_fn=<NllLossBackward0>)
9900 tensor(5.2532, device='cuda:0', grad_fn=<NllLossBackward0>)
98000
10000 tensor(5.4727, device='cuda:0', grad_fn=<NllLossBackward0>)
99000
10100 tensor(5.3607, device='cuda:0', grad_fn=<NllLossBackward0>)
100000
10200 tensor(5.2989, device='cuda:0', grad_fn=<NllLossBackward0>)
101000
10300 tensor(5.4168, device='cuda:0', grad_fn=<NllLossBackward0>)
102000
10400 tensor(5.4272, device='cuda:0', grad_fn=<NllLossBackward0>)
103000
10500 tensor(5.4838, device='cuda:0', grad_fn=<NllLossBackward0>)
104000
10600 tensor(5.5675, device='cuda:0', grad_fn=<NllLossBackward0>)
105000
10700 tensor(5.4027, device='cuda:0', grad_fn=<NllLossBackward0>)
106000
10800 tensor(5.4252, device='cuda:0', grad_fn=<NllLossBackward0>)
107000
10900 tensor(5.3408, device='cuda:0', grad_fn=<NllLossBackward0>)
108000
11000 tensor(5.5754, device='cuda:0', grad_fn=<NllLossBackward0>)
109000
11100 tensor(5.1920, device='cuda:0', grad_fn=<NllLossBackward0>)
110000
11200 tensor(5.3604, device='cuda:0', grad_fn=<NllLossBackward0>)
111000
11300 tensor(5.3836, device='cuda:0', grad_fn=<NllLossBackward0>)
112000
11400 tensor(5.3330, device='cuda:0', grad_fn=<NllLossBackward0>)
113000
11500 tensor(5.4023, device='cuda:0', grad_fn=<NllLossBackward0>)
114000
11600 tensor(5.3923, device='cuda:0', grad_fn=<NllLossBackward0>)
115000
11700 tensor(5.3145, device='cuda:0', grad_fn=<NllLossBackward0>)
116000
11800 tensor(5.5174, device='cuda:0', grad_fn=<NllLossBackward0>)
117000
11900 tensor(5.3522, device='cuda:0', grad_fn=<NllLossBackward0>)
118000
12000 tensor(5.4232, device='cuda:0', grad_fn=<NllLossBackward0>)
119000
12100 tensor(5.4382, device='cuda:0', grad_fn=<NllLossBackward0>)
120000
12200 tensor(5.4488, device='cuda:0', grad_fn=<NllLossBackward0>)
121000
12300 tensor(5.5409, device='cuda:0', grad_fn=<NllLossBackward0>)
122000
12400 tensor(5.4200, device='cuda:0', grad_fn=<NllLossBackward0>)
123000
12500 tensor(5.3292, device='cuda:0', grad_fn=<NllLossBackward0>)
124000
12600 tensor(5.3788, device='cuda:0', grad_fn=<NllLossBackward0>)
125000
12700 tensor(5.3116, device='cuda:0', grad_fn=<NllLossBackward0>)
126000
12800 tensor(5.4948, device='cuda:0', grad_fn=<NllLossBackward0>)
127000
12900 tensor(5.3557, device='cuda:0', grad_fn=<NllLossBackward0>)
128000
13000 tensor(5.1732, device='cuda:0', grad_fn=<NllLossBackward0>)
129000
13100 tensor(5.3782, device='cuda:0', grad_fn=<NllLossBackward0>)
130000
13200 tensor(5.4178, device='cuda:0', grad_fn=<NllLossBackward0>)
131000
13300 tensor(5.2929, device='cuda:0', grad_fn=<NllLossBackward0>)
132000
13400 tensor(5.3806, device='cuda:0', grad_fn=<NllLossBackward0>)
133000
13500 tensor(5.3394, device='cuda:0', grad_fn=<NllLossBackward0>)
134000
13600 tensor(5.4191, device='cuda:0', grad_fn=<NllLossBackward0>)
135000
13700 tensor(5.3856, device='cuda:0', grad_fn=<NllLossBackward0>)
136000
13800 tensor(5.3839, device='cuda:0', grad_fn=<NllLossBackward0>)
137000
13900 tensor(5.2391, device='cuda:0', grad_fn=<NllLossBackward0>)
138000
14000 tensor(5.4865, device='cuda:0', grad_fn=<NllLossBackward0>)
139000
14100 tensor(5.1952, device='cuda:0', grad_fn=<NllLossBackward0>)
140000
14200 tensor(5.4670, device='cuda:0', grad_fn=<NllLossBackward0>)
141000
14300 tensor(5.4385, device='cuda:0', grad_fn=<NllLossBackward0>)
142000
14400 tensor(5.3347, device='cuda:0', grad_fn=<NllLossBackward0>)
143000
14500 tensor(5.4370, device='cuda:0', grad_fn=<NllLossBackward0>)
144000
14600 tensor(5.4695, device='cuda:0', grad_fn=<NllLossBackward0>)
145000
14700 tensor(5.3453, device='cuda:0', grad_fn=<NllLossBackward0>)
146000
14800 tensor(5.7928, device='cuda:0', grad_fn=<NllLossBackward0>)
147000
14900 tensor(5.4451, device='cuda:0', grad_fn=<NllLossBackward0>)
148000
15000 tensor(5.3087, device='cuda:0', grad_fn=<NllLossBackward0>)
149000
15100 tensor(5.5241, device='cuda:0', grad_fn=<NllLossBackward0>)
150000
15200 tensor(5.3894, device='cuda:0', grad_fn=<NllLossBackward0>)
151000
15300 tensor(5.3809, device='cuda:0', grad_fn=<NllLossBackward0>)
152000
15400 tensor(5.2696, device='cuda:0', grad_fn=<NllLossBackward0>)
153000
15500 tensor(5.4343, device='cuda:0', grad_fn=<NllLossBackward0>)
154000
15600 tensor(5.4322, device='cuda:0', grad_fn=<NllLossBackward0>)
155000
15700 tensor(5.3296, device='cuda:0', grad_fn=<NllLossBackward0>)
156000
15800 tensor(5.2456, device='cuda:0', grad_fn=<NllLossBackward0>)
157000
15900 tensor(5.3806, device='cuda:0', grad_fn=<NllLossBackward0>)
158000
16000 tensor(5.2008, device='cuda:0', grad_fn=<NllLossBackward0>)
159000
16100 tensor(5.2489, device='cuda:0', grad_fn=<NllLossBackward0>)
160000
16200 tensor(5.5902, device='cuda:0', grad_fn=<NllLossBackward0>)
161000
16300 tensor(5.4159, device='cuda:0', grad_fn=<NllLossBackward0>)
162000
16400 tensor(5.3966, device='cuda:0', grad_fn=<NllLossBackward0>)
163000
16500 tensor(5.5113, device='cuda:0', grad_fn=<NllLossBackward0>)
164000
16600 tensor(5.3599, device='cuda:0', grad_fn=<NllLossBackward0>)
165000
16700 tensor(5.3372, device='cuda:0', grad_fn=<NllLossBackward0>)
166000
16800 tensor(5.4158, device='cuda:0', grad_fn=<NllLossBackward0>)
167000
16900 tensor(5.1788, device='cuda:0', grad_fn=<NllLossBackward0>)
168000
17000 tensor(5.4497, device='cuda:0', grad_fn=<NllLossBackward0>)
169000
17100 tensor(5.2981, device='cuda:0', grad_fn=<NllLossBackward0>)
170000
17200 tensor(5.4330, device='cuda:0', grad_fn=<NllLossBackward0>)
171000
17300 tensor(5.4495, device='cuda:0', grad_fn=<NllLossBackward0>)
172000
17400 tensor(5.2431, device='cuda:0', grad_fn=<NllLossBackward0>)
173000
17500 tensor(5.2652, device='cuda:0', grad_fn=<NllLossBackward0>)
174000
17600 tensor(5.3007, device='cuda:0', grad_fn=<NllLossBackward0>)
175000
17700 tensor(5.2852, device='cuda:0', grad_fn=<NllLossBackward0>)
176000
17800 tensor(5.3431, device='cuda:0', grad_fn=<NllLossBackward0>)
177000
17900 tensor(5.2395, device='cuda:0', grad_fn=<NllLossBackward0>)
178000
18000 tensor(5.4841, device='cuda:0', grad_fn=<NllLossBackward0>)
179000
18100 tensor(5.4218, device='cuda:0', grad_fn=<NllLossBackward0>)
180000
18200 tensor(5.3397, device='cuda:0', grad_fn=<NllLossBackward0>)
181000
18300 tensor(5.3426, device='cuda:0', grad_fn=<NllLossBackward0>)
182000
18400 tensor(5.3654, device='cuda:0', grad_fn=<NllLossBackward0>)
183000
18500 tensor(5.3484, device='cuda:0', grad_fn=<NllLossBackward0>)
184000
18600 tensor(5.5509, device='cuda:0', grad_fn=<NllLossBackward0>)
185000
18700 tensor(5.3702, device='cuda:0', grad_fn=<NllLossBackward0>)
186000
18800 tensor(5.5361, device='cuda:0', grad_fn=<NllLossBackward0>)
187000
18900 tensor(5.4132, device='cuda:0', grad_fn=<NllLossBackward0>)
188000
19000 tensor(5.4235, device='cuda:0', grad_fn=<NllLossBackward0>)
189000
19100 tensor(5.5318, device='cuda:0', grad_fn=<NllLossBackward0>)
190000
19200 tensor(5.4136, device='cuda:0', grad_fn=<NllLossBackward0>)
191000
19300 tensor(5.5053, device='cuda:0', grad_fn=<NllLossBackward0>)
192000
19400 tensor(5.3472, device='cuda:0', grad_fn=<NllLossBackward0>)
193000
19500 tensor(5.3511, device='cuda:0', grad_fn=<NllLossBackward0>)
194000
19600 tensor(5.3861, device='cuda:0', grad_fn=<NllLossBackward0>)
195000
19700 tensor(5.4345, device='cuda:0', grad_fn=<NllLossBackward0>)
196000
19800 tensor(5.3067, device='cuda:0', grad_fn=<NllLossBackward0>)
197000
19900 tensor(5.3079, device='cuda:0', grad_fn=<NllLossBackward0>)
198000
20000 tensor(5.3268, device='cuda:0', grad_fn=<NllLossBackward0>)
199000
20100 tensor(5.2668, device='cuda:0', grad_fn=<NllLossBackward0>)
200000
20200 tensor(5.1998, device='cuda:0', grad_fn=<NllLossBackward0>)
201000
20300 tensor(5.3105, device='cuda:0', grad_fn=<NllLossBackward0>)
20400 tensor(5.3584, device='cuda:0', grad_fn=<NllLossBackward0>)
202000
20500 tensor(5.3580, device='cuda:0', grad_fn=<NllLossBackward0>)
203000
20600 tensor(5.5528, device='cuda:0', grad_fn=<NllLossBackward0>)
204000
20700 tensor(5.3871, device='cuda:0', grad_fn=<NllLossBackward0>)
205000
20800 tensor(5.2208, device='cuda:0', grad_fn=<NllLossBackward0>)
206000
20900 tensor(5.5007, device='cuda:0', grad_fn=<NllLossBackward0>)
207000
21000 tensor(5.3396, device='cuda:0', grad_fn=<NllLossBackward0>)
208000
21100 tensor(5.3407, device='cuda:0', grad_fn=<NllLossBackward0>)
209000
21200 tensor(5.2243, device='cuda:0', grad_fn=<NllLossBackward0>)
210000
21300 tensor(5.4206, device='cuda:0', grad_fn=<NllLossBackward0>)
211000
21400 tensor(5.4574, device='cuda:0', grad_fn=<NllLossBackward0>)
212000
21500 tensor(5.2328, device='cuda:0', grad_fn=<NllLossBackward0>)
213000
21600 tensor(5.2233, device='cuda:0', grad_fn=<NllLossBackward0>)
214000
21700 tensor(5.2152, device='cuda:0', grad_fn=<NllLossBackward0>)
215000
21800 tensor(5.3497, device='cuda:0', grad_fn=<NllLossBackward0>)
216000
21900 tensor(5.3425, device='cuda:0', grad_fn=<NllLossBackward0>)
217000
22000 tensor(5.3277, device='cuda:0', grad_fn=<NllLossBackward0>)
218000
22100 tensor(5.2012, device='cuda:0', grad_fn=<NllLossBackward0>)
219000
22200 tensor(5.0736, device='cuda:0', grad_fn=<NllLossBackward0>)
220000
22300 tensor(5.5070, device='cuda:0', grad_fn=<NllLossBackward0>)
221000
22400 tensor(5.2190, device='cuda:0', grad_fn=<NllLossBackward0>)
222000
22500 tensor(5.2434, device='cuda:0', grad_fn=<NllLossBackward0>)
223000
22600 tensor(5.4325, device='cuda:0', grad_fn=<NllLossBackward0>)
224000
22700 tensor(5.1909, device='cuda:0', grad_fn=<NllLossBackward0>)
225000
22800 tensor(5.4576, device='cuda:0', grad_fn=<NllLossBackward0>)
226000
22900 tensor(5.5069, device='cuda:0', grad_fn=<NllLossBackward0>)
227000
23000 tensor(5.4041, device='cuda:0', grad_fn=<NllLossBackward0>)
228000
23100 tensor(5.3908, device='cuda:0', grad_fn=<NllLossBackward0>)
229000
23200 tensor(5.3866, device='cuda:0', grad_fn=<NllLossBackward0>)
230000
23300 tensor(5.4714, device='cuda:0', grad_fn=<NllLossBackward0>)
231000
23400 tensor(5.4781, device='cuda:0', grad_fn=<NllLossBackward0>)
232000
23500 tensor(5.3154, device='cuda:0', grad_fn=<NllLossBackward0>)
233000
23600 tensor(5.2854, device='cuda:0', grad_fn=<NllLossBackward0>)
234000
23700 tensor(5.3050, device='cuda:0', grad_fn=<NllLossBackward0>)
235000
23800 tensor(5.1721, device='cuda:0', grad_fn=<NllLossBackward0>)
236000
23900 tensor(5.2637, device='cuda:0', grad_fn=<NllLossBackward0>)
237000
24000 tensor(5.2519, device='cuda:0', grad_fn=<NllLossBackward0>)
238000
24100 tensor(5.3407, device='cuda:0', grad_fn=<NllLossBackward0>)
239000
24200 tensor(5.5137, device='cuda:0', grad_fn=<NllLossBackward0>)
240000
24300 tensor(5.4080, device='cuda:0', grad_fn=<NllLossBackward0>)
241000
24400 tensor(5.5379, device='cuda:0', grad_fn=<NllLossBackward0>)
242000
24500 tensor(5.3255, device='cuda:0', grad_fn=<NllLossBackward0>)
243000
24600 tensor(5.4515, device='cuda:0', grad_fn=<NllLossBackward0>)
244000
24700 tensor(5.3535, device='cuda:0', grad_fn=<NllLossBackward0>)
245000
24800 tensor(5.3935, device='cuda:0', grad_fn=<NllLossBackward0>)
246000
24900 tensor(5.4553, device='cuda:0', grad_fn=<NllLossBackward0>)
247000
25000 tensor(5.4708, device='cuda:0', grad_fn=<NllLossBackward0>)
248000
25100 tensor(5.3920, device='cuda:0', grad_fn=<NllLossBackward0>)
249000
25200 tensor(5.4083, device='cuda:0', grad_fn=<NllLossBackward0>)
250000
25300 tensor(5.4332, device='cuda:0', grad_fn=<NllLossBackward0>)
251000
25400 tensor(5.4136, device='cuda:0', grad_fn=<NllLossBackward0>)
252000
25500 tensor(5.3147, device='cuda:0', grad_fn=<NllLossBackward0>)
253000
25600 tensor(5.5860, device='cuda:0', grad_fn=<NllLossBackward0>)
254000
25700 tensor(5.3490, device='cuda:0', grad_fn=<NllLossBackward0>)
255000
25800 tensor(5.4464, device='cuda:0', grad_fn=<NllLossBackward0>)
256000
25900 tensor(5.3857, device='cuda:0', grad_fn=<NllLossBackward0>)
257000
26000 tensor(5.3893, device='cuda:0', grad_fn=<NllLossBackward0>)
258000
26100 tensor(5.3041, device='cuda:0', grad_fn=<NllLossBackward0>)
259000
26200 tensor(5.2321, device='cuda:0', grad_fn=<NllLossBackward0>)
260000
26300 tensor(5.4289, device='cuda:0', grad_fn=<NllLossBackward0>)
261000
26400 tensor(5.4663, device='cuda:0', grad_fn=<NllLossBackward0>)
262000
26500 tensor(5.1922, device='cuda:0', grad_fn=<NllLossBackward0>)
263000
26600 tensor(5.5283, device='cuda:0', grad_fn=<NllLossBackward0>)
264000
26700 tensor(5.3933, device='cuda:0', grad_fn=<NllLossBackward0>)
265000
26800 tensor(5.5680, device='cuda:0', grad_fn=<NllLossBackward0>)
266000
26900 tensor(5.3281, device='cuda:0', grad_fn=<NllLossBackward0>)
267000
27000 tensor(5.2408, device='cuda:0', grad_fn=<NllLossBackward0>)
268000
27100 tensor(5.2671, device='cuda:0', grad_fn=<NllLossBackward0>)
269000
27200 tensor(5.3099, device='cuda:0', grad_fn=<NllLossBackward0>)
270000
27300 tensor(5.5049, device='cuda:0', grad_fn=<NllLossBackward0>)
271000
27400 tensor(5.3850, device='cuda:0', grad_fn=<NllLossBackward0>)
272000
27500 tensor(5.2843, device='cuda:0', grad_fn=<NllLossBackward0>)
273000
27600 tensor(5.5777, device='cuda:0', grad_fn=<NllLossBackward0>)
274000
27700 tensor(5.4017, device='cuda:0', grad_fn=<NllLossBackward0>)
275000
27800 tensor(5.3994, device='cuda:0', grad_fn=<NllLossBackward0>)
276000
27900 tensor(5.5128, device='cuda:0', grad_fn=<NllLossBackward0>)
277000
28000 tensor(5.3708, device='cuda:0', grad_fn=<NllLossBackward0>)
278000
28100 tensor(5.3382, device='cuda:0', grad_fn=<NllLossBackward0>)
279000
28200 tensor(5.4996, device='cuda:0', grad_fn=<NllLossBackward0>)
280000
28300 tensor(5.1214, device='cuda:0', grad_fn=<NllLossBackward0>)
281000
28400 tensor(5.5647, device='cuda:0', grad_fn=<NllLossBackward0>)
282000
28500 tensor(5.3959, device='cuda:0', grad_fn=<NllLossBackward0>)
283000
28600 tensor(5.3312, device='cuda:0', grad_fn=<NllLossBackward0>)
284000
28700 tensor(5.4663, device='cuda:0', grad_fn=<NllLossBackward0>)
285000
28800 tensor(5.5155, device='cuda:0', grad_fn=<NllLossBackward0>)
286000
28900 tensor(5.3872, device='cuda:0', grad_fn=<NllLossBackward0>)
287000
29000 tensor(5.3017, device='cuda:0', grad_fn=<NllLossBackward0>)
288000
29100 tensor(5.0583, device='cuda:0', grad_fn=<NllLossBackward0>)
289000
29200 tensor(5.2099, device='cuda:0', grad_fn=<NllLossBackward0>)
290000
29300 tensor(5.4934, device='cuda:0', grad_fn=<NllLossBackward0>)
291000
29400 tensor(5.6202, device='cuda:0', grad_fn=<NllLossBackward0>)
292000
29500 tensor(5.4016, device='cuda:0', grad_fn=<NllLossBackward0>)
293000
29600 tensor(5.2601, device='cuda:0', grad_fn=<NllLossBackward0>)
294000
29700 tensor(5.4038, device='cuda:0', grad_fn=<NllLossBackward0>)
295000
29800 tensor(5.2475, device='cuda:0', grad_fn=<NllLossBackward0>)
296000
29900 tensor(5.4960, device='cuda:0', grad_fn=<NllLossBackward0>)
297000
30000 tensor(5.2438, device='cuda:0', grad_fn=<NllLossBackward0>)
298000
30100 tensor(5.3221, device='cuda:0', grad_fn=<NllLossBackward0>)
299000
30200 tensor(5.2686, device='cuda:0', grad_fn=<NllLossBackward0>)
300000
30300 tensor(5.3735, device='cuda:0', grad_fn=<NllLossBackward0>)
301000
30400 tensor(5.2057, device='cuda:0', grad_fn=<NllLossBackward0>)
302000
30500 tensor(5.3767, device='cuda:0', grad_fn=<NllLossBackward0>)
30600 tensor(5.3515, device='cuda:0', grad_fn=<NllLossBackward0>)
303000
30700 tensor(5.3841, device='cuda:0', grad_fn=<NllLossBackward0>)
304000
30800 tensor(5.3889, device='cuda:0', grad_fn=<NllLossBackward0>)
305000
30900 tensor(5.4117, device='cuda:0', grad_fn=<NllLossBackward0>)
306000
31000 tensor(5.5205, device='cuda:0', grad_fn=<NllLossBackward0>)
307000
31100 tensor(5.1742, device='cuda:0', grad_fn=<NllLossBackward0>)
308000
31200 tensor(5.2173, device='cuda:0', grad_fn=<NllLossBackward0>)
309000
31300 tensor(5.4785, device='cuda:0', grad_fn=<NllLossBackward0>)
310000
31400 tensor(5.2577, device='cuda:0', grad_fn=<NllLossBackward0>)
311000
31500 tensor(5.4429, device='cuda:0', grad_fn=<NllLossBackward0>)
312000
31600 tensor(5.4289, device='cuda:0', grad_fn=<NllLossBackward0>)
313000
31700 tensor(5.3961, device='cuda:0', grad_fn=<NllLossBackward0>)
314000
31800 tensor(5.4999, device='cuda:0', grad_fn=<NllLossBackward0>)
315000
31900 tensor(5.1248, device='cuda:0', grad_fn=<NllLossBackward0>)
316000
32000 tensor(5.3122, device='cuda:0', grad_fn=<NllLossBackward0>)
317000
32100 tensor(5.1931, device='cuda:0', grad_fn=<NllLossBackward0>)
318000
32200 tensor(5.5096, device='cuda:0', grad_fn=<NllLossBackward0>)
319000
32300 tensor(5.4973, device='cuda:0', grad_fn=<NllLossBackward0>)
320000
32400 tensor(5.4742, device='cuda:0', grad_fn=<NllLossBackward0>)
321000
32500 tensor(5.2964, device='cuda:0', grad_fn=<NllLossBackward0>)
322000
32600 tensor(5.4063, device='cuda:0', grad_fn=<NllLossBackward0>)
323000
32700 tensor(5.3369, device='cuda:0', grad_fn=<NllLossBackward0>)
324000
32800 tensor(5.5636, device='cuda:0', grad_fn=<NllLossBackward0>)
325000
32900 tensor(5.4245, device='cuda:0', grad_fn=<NllLossBackward0>)
326000
33000 tensor(5.2032, device='cuda:0', grad_fn=<NllLossBackward0>)
327000
33100 tensor(5.4095, device='cuda:0', grad_fn=<NllLossBackward0>)
328000
33200 tensor(5.5071, device='cuda:0', grad_fn=<NllLossBackward0>)
329000
33300 tensor(5.2729, device='cuda:0', grad_fn=<NllLossBackward0>)
330000
33400 tensor(5.5492, device='cuda:0', grad_fn=<NllLossBackward0>)
331000
33500 tensor(5.3701, device='cuda:0', grad_fn=<NllLossBackward0>)
332000
33600 tensor(5.3223, device='cuda:0', grad_fn=<NllLossBackward0>)
333000
33700 tensor(5.3725, device='cuda:0', grad_fn=<NllLossBackward0>)
334000
33800 tensor(5.4572, device='cuda:0', grad_fn=<NllLossBackward0>)
335000
33900 tensor(5.1889, device='cuda:0', grad_fn=<NllLossBackward0>)
336000
34000 tensor(5.4090, device='cuda:0', grad_fn=<NllLossBackward0>)
337000
34100 tensor(5.3798, device='cuda:0', grad_fn=<NllLossBackward0>)
338000
34200 tensor(5.4259, device='cuda:0', grad_fn=<NllLossBackward0>)
339000
34300 tensor(5.2132, device='cuda:0', grad_fn=<NllLossBackward0>)
340000
34400 tensor(5.6692, device='cuda:0', grad_fn=<NllLossBackward0>)
341000
34500 tensor(5.5324, device='cuda:0', grad_fn=<NllLossBackward0>)
342000
34600 tensor(5.4271, device='cuda:0', grad_fn=<NllLossBackward0>)
343000
34700 tensor(5.4978, device='cuda:0', grad_fn=<NllLossBackward0>)
344000
34800 tensor(5.5230, device='cuda:0', grad_fn=<NllLossBackward0>)
345000
34900 tensor(5.5652, device='cuda:0', grad_fn=<NllLossBackward0>)
346000
35000 tensor(5.5478, device='cuda:0', grad_fn=<NllLossBackward0>)
347000
35100 tensor(5.3700, device='cuda:0', grad_fn=<NllLossBackward0>)
348000
35200 tensor(5.2958, device='cuda:0', grad_fn=<NllLossBackward0>)
349000
35300 tensor(5.5219, device='cuda:0', grad_fn=<NllLossBackward0>)
350000
35400 tensor(5.1702, device='cuda:0', grad_fn=<NllLossBackward0>)
351000
35500 tensor(5.2604, device='cuda:0', grad_fn=<NllLossBackward0>)
352000
35600 tensor(5.3821, device='cuda:0', grad_fn=<NllLossBackward0>)
353000
35700 tensor(5.2551, device='cuda:0', grad_fn=<NllLossBackward0>)
354000
35800 tensor(5.3840, device='cuda:0', grad_fn=<NllLossBackward0>)
355000
35900 tensor(5.3635, device='cuda:0', grad_fn=<NllLossBackward0>)
356000
36000 tensor(5.1400, device='cuda:0', grad_fn=<NllLossBackward0>)
357000
36100 tensor(5.5134, device='cuda:0', grad_fn=<NllLossBackward0>)
358000
36200 tensor(5.3632, device='cuda:0', grad_fn=<NllLossBackward0>)
359000
36300 tensor(5.6461, device='cuda:0', grad_fn=<NllLossBackward0>)
360000
36400 tensor(5.3415, device='cuda:0', grad_fn=<NllLossBackward0>)
361000
36500 tensor(5.3659, device='cuda:0', grad_fn=<NllLossBackward0>)
362000
36600 tensor(5.3874, device='cuda:0', grad_fn=<NllLossBackward0>)
363000
36700 tensor(5.1886, device='cuda:0', grad_fn=<NllLossBackward0>)
364000
36800 tensor(5.2958, device='cuda:0', grad_fn=<NllLossBackward0>)
365000
36900 tensor(5.4094, device='cuda:0', grad_fn=<NllLossBackward0>)
366000
37000 tensor(5.3023, device='cuda:0', grad_fn=<NllLossBackward0>)
367000
37100 tensor(5.3287, device='cuda:0', grad_fn=<NllLossBackward0>)
368000
37200 tensor(5.3996, device='cuda:0', grad_fn=<NllLossBackward0>)
369000
37300 tensor(5.3001, device='cuda:0', grad_fn=<NllLossBackward0>)
370000
37400 tensor(5.6516, device='cuda:0', grad_fn=<NllLossBackward0>)
371000
37500 tensor(5.3366, device='cuda:0', grad_fn=<NllLossBackward0>)
372000
37600 tensor(5.3282, device='cuda:0', grad_fn=<NllLossBackward0>)
373000
37700 tensor(5.5061, device='cuda:0', grad_fn=<NllLossBackward0>)
374000
37800 tensor(5.3408, device='cuda:0', grad_fn=<NllLossBackward0>)
375000
37900 tensor(5.3203, device='cuda:0', grad_fn=<NllLossBackward0>)
376000
38000 tensor(5.3996, device='cuda:0', grad_fn=<NllLossBackward0>)
377000
38100 tensor(5.4133, device='cuda:0', grad_fn=<NllLossBackward0>)
378000
38200 tensor(5.4262, device='cuda:0', grad_fn=<NllLossBackward0>)
379000
38300 tensor(5.3305, device='cuda:0', grad_fn=<NllLossBackward0>)
380000
38400 tensor(5.3983, device='cuda:0', grad_fn=<NllLossBackward0>)
381000
38500 tensor(5.4246, device='cuda:0', grad_fn=<NllLossBackward0>)
382000
38600 tensor(5.3713, device='cuda:0', grad_fn=<NllLossBackward0>)
383000
38700 tensor(5.3634, device='cuda:0', grad_fn=<NllLossBackward0>)
384000
38800 tensor(5.4504, device='cuda:0', grad_fn=<NllLossBackward0>)
385000
38900 tensor(5.5273, device='cuda:0', grad_fn=<NllLossBackward0>)
386000
39000 tensor(5.2229, device='cuda:0', grad_fn=<NllLossBackward0>)
387000
39100 tensor(5.4503, device='cuda:0', grad_fn=<NllLossBackward0>)
388000
39200 tensor(5.5406, device='cuda:0', grad_fn=<NllLossBackward0>)
389000
39300 tensor(5.3640, device='cuda:0', grad_fn=<NllLossBackward0>)
390000
39400 tensor(5.4311, device='cuda:0', grad_fn=<NllLossBackward0>)
391000
39500 tensor(5.5292, device='cuda:0', grad_fn=<NllLossBackward0>)
392000
39600 tensor(5.2217, device='cuda:0', grad_fn=<NllLossBackward0>)
393000
39700 tensor(5.2121, device='cuda:0', grad_fn=<NllLossBackward0>)
394000
39800 tensor(5.3415, device='cuda:0', grad_fn=<NllLossBackward0>)
395000
39900 tensor(5.1605, device='cuda:0', grad_fn=<NllLossBackward0>)
396000
40000 tensor(5.2472, device='cuda:0', grad_fn=<NllLossBackward0>)
397000
40100 tensor(5.3351, device='cuda:0', grad_fn=<NllLossBackward0>)
398000
40200 tensor(5.3198, device='cuda:0', grad_fn=<NllLossBackward0>)
399000
40300 tensor(5.3862, device='cuda:0', grad_fn=<NllLossBackward0>)
40400 tensor(5.3946, device='cuda:0', grad_fn=<NllLossBackward0>)
400000
40500 tensor(5.3120, device='cuda:0', grad_fn=<NllLossBackward0>)
401000
40600 tensor(5.3741, device='cuda:0', grad_fn=<NllLossBackward0>)
402000
40700 tensor(5.4199, device='cuda:0', grad_fn=<NllLossBackward0>)
403000
40800 tensor(5.3702, device='cuda:0', grad_fn=<NllLossBackward0>)
404000
40900 tensor(5.3212, device='cuda:0', grad_fn=<NllLossBackward0>)
405000
41000 tensor(5.3683, device='cuda:0', grad_fn=<NllLossBackward0>)
406000
41100 tensor(5.3491, device='cuda:0', grad_fn=<NllLossBackward0>)
407000
41200 tensor(5.2400, device='cuda:0', grad_fn=<NllLossBackward0>)
408000
41300 tensor(5.3728, device='cuda:0', grad_fn=<NllLossBackward0>)
409000
41400 tensor(5.2643, device='cuda:0', grad_fn=<NllLossBackward0>)
410000
41500 tensor(5.4064, device='cuda:0', grad_fn=<NllLossBackward0>)
411000
41600 tensor(5.3238, device='cuda:0', grad_fn=<NllLossBackward0>)
412000
41700 tensor(5.3469, device='cuda:0', grad_fn=<NllLossBackward0>)
413000
41800 tensor(5.3432, device='cuda:0', grad_fn=<NllLossBackward0>)
414000
41900 tensor(5.3521, device='cuda:0', grad_fn=<NllLossBackward0>)
415000
42000 tensor(5.5087, device='cuda:0', grad_fn=<NllLossBackward0>)
416000
42100 tensor(5.2556, device='cuda:0', grad_fn=<NllLossBackward0>)
417000
42200 tensor(5.3407, device='cuda:0', grad_fn=<NllLossBackward0>)
418000
42300 tensor(5.4058, device='cuda:0', grad_fn=<NllLossBackward0>)
419000
42400 tensor(5.2231, device='cuda:0', grad_fn=<NllLossBackward0>)
420000
42500 tensor(5.3912, device='cuda:0', grad_fn=<NllLossBackward0>)
421000
42600 tensor(5.1878, device='cuda:0', grad_fn=<NllLossBackward0>)
422000
42700 tensor(5.4955, device='cuda:0', grad_fn=<NllLossBackward0>)
423000
42800 tensor(5.4193, device='cuda:0', grad_fn=<NllLossBackward0>)
424000
42900 tensor(5.2662, device='cuda:0', grad_fn=<NllLossBackward0>)
425000
43000 tensor(5.4093, device='cuda:0', grad_fn=<NllLossBackward0>)
426000
43100 tensor(5.4089, device='cuda:0', grad_fn=<NllLossBackward0>)
427000
43200 tensor(5.2223, device='cuda:0', grad_fn=<NllLossBackward0>)
428000
43300 tensor(5.2456, device='cuda:0', grad_fn=<NllLossBackward0>)
429000
43400 tensor(5.4129, device='cuda:0', grad_fn=<NllLossBackward0>)
430000
43500 tensor(5.1283, device='cuda:0', grad_fn=<NllLossBackward0>)
431000
43600 tensor(5.3275, device='cuda:0', grad_fn=<NllLossBackward0>)
432000
epoch: = 4
0 tensor(5.3172, device='cuda:0', grad_fn=<NllLossBackward0>)
1000
100 tensor(5.3864, device='cuda:0', grad_fn=<NllLossBackward0>)
200 tensor(5.2618, device='cuda:0', grad_fn=<NllLossBackward0>)
2000
300 tensor(5.2652, device='cuda:0', grad_fn=<NllLossBackward0>)
3000
400 tensor(5.2749, device='cuda:0', grad_fn=<NllLossBackward0>)
4000
500 tensor(5.4347, device='cuda:0', grad_fn=<NllLossBackward0>)
5000
600 tensor(5.2271, device='cuda:0', grad_fn=<NllLossBackward0>)
6000
700 tensor(5.5396, device='cuda:0', grad_fn=<NllLossBackward0>)
7000
800 tensor(5.1379, device='cuda:0', grad_fn=<NllLossBackward0>)
8000
900 tensor(5.3861, device='cuda:0', grad_fn=<NllLossBackward0>)
9000
1000 tensor(5.2629, device='cuda:0', grad_fn=<NllLossBackward0>)
10000
1100 tensor(5.4575, device='cuda:0', grad_fn=<NllLossBackward0>)
11000
1200 tensor(5.4936, device='cuda:0', grad_fn=<NllLossBackward0>)
12000
1300 tensor(5.4281, device='cuda:0', grad_fn=<NllLossBackward0>)
13000
1400 tensor(5.4186, device='cuda:0', grad_fn=<NllLossBackward0>)
14000
1500 tensor(5.5070, device='cuda:0', grad_fn=<NllLossBackward0>)
15000
1600 tensor(5.1769, device='cuda:0', grad_fn=<NllLossBackward0>)
16000
1700 tensor(5.2856, device='cuda:0', grad_fn=<NllLossBackward0>)
17000
1800 tensor(5.2827, device='cuda:0', grad_fn=<NllLossBackward0>)
18000
1900 tensor(5.2544, device='cuda:0', grad_fn=<NllLossBackward0>)
19000
2000 tensor(5.3218, device='cuda:0', grad_fn=<NllLossBackward0>)
20000
2100 tensor(5.4549, device='cuda:0', grad_fn=<NllLossBackward0>)
21000
2200 tensor(5.2864, device='cuda:0', grad_fn=<NllLossBackward0>)
22000
2300 tensor(5.3145, device='cuda:0', grad_fn=<NllLossBackward0>)
23000
2400 tensor(5.2987, device='cuda:0', grad_fn=<NllLossBackward0>)
24000
2500 tensor(5.3498, device='cuda:0', grad_fn=<NllLossBackward0>)
25000
2600 tensor(5.3730, device='cuda:0', grad_fn=<NllLossBackward0>)
26000
2700 tensor(5.4017, device='cuda:0', grad_fn=<NllLossBackward0>)
27000
2800 tensor(5.4255, device='cuda:0', grad_fn=<NllLossBackward0>)
28000
2900 tensor(5.5475, device='cuda:0', grad_fn=<NllLossBackward0>)
29000
3000 tensor(5.2988, device='cuda:0', grad_fn=<NllLossBackward0>)
30000
3100 tensor(5.3753, device='cuda:0', grad_fn=<NllLossBackward0>)
31000
3200 tensor(5.4049, device='cuda:0', grad_fn=<NllLossBackward0>)
32000
3300 tensor(5.3206, device='cuda:0', grad_fn=<NllLossBackward0>)
33000
3400 tensor(5.2159, device='cuda:0', grad_fn=<NllLossBackward0>)
34000
3500 tensor(5.3423, device='cuda:0', grad_fn=<NllLossBackward0>)
35000
3600 tensor(5.3717, device='cuda:0', grad_fn=<NllLossBackward0>)
36000
3700 tensor(5.3042, device='cuda:0', grad_fn=<NllLossBackward0>)
37000
3800 tensor(5.3258, device='cuda:0', grad_fn=<NllLossBackward0>)
38000
3900 tensor(5.1989, device='cuda:0', grad_fn=<NllLossBackward0>)
39000
4000 tensor(5.2650, device='cuda:0', grad_fn=<NllLossBackward0>)
40000
4100 tensor(5.3953, device='cuda:0', grad_fn=<NllLossBackward0>)
41000
4200 tensor(5.4542, device='cuda:0', grad_fn=<NllLossBackward0>)
42000
4300 tensor(5.3466, device='cuda:0', grad_fn=<NllLossBackward0>)
43000
4400 tensor(5.4222, device='cuda:0', grad_fn=<NllLossBackward0>)
44000
4500 tensor(5.2254, device='cuda:0', grad_fn=<NllLossBackward0>)
45000
4600 tensor(5.5610, device='cuda:0', grad_fn=<NllLossBackward0>)
46000
4700 tensor(5.2753, device='cuda:0', grad_fn=<NllLossBackward0>)
47000
4800 tensor(5.4028, device='cuda:0', grad_fn=<NllLossBackward0>)
48000
4900 tensor(5.4516, device='cuda:0', grad_fn=<NllLossBackward0>)
49000
5000 tensor(5.4464, device='cuda:0', grad_fn=<NllLossBackward0>)
50000
5100 tensor(5.5018, device='cuda:0', grad_fn=<NllLossBackward0>)
51000
5200 tensor(5.5194, device='cuda:0', grad_fn=<NllLossBackward0>)
52000
5300 tensor(5.5077, device='cuda:0', grad_fn=<NllLossBackward0>)
53000
5400 tensor(5.3746, device='cuda:0', grad_fn=<NllLossBackward0>)
54000
5500 tensor(5.4847, device='cuda:0', grad_fn=<NllLossBackward0>)
55000
5600 tensor(5.2664, device='cuda:0', grad_fn=<NllLossBackward0>)
56000
5700 tensor(5.5265, device='cuda:0', grad_fn=<NllLossBackward0>)
57000
5800 tensor(5.5101, device='cuda:0', grad_fn=<NllLossBackward0>)
58000
5900 tensor(5.4513, device='cuda:0', grad_fn=<NllLossBackward0>)
59000
6000 tensor(5.3554, device='cuda:0', grad_fn=<NllLossBackward0>)
60000
6100 tensor(5.3616, device='cuda:0', grad_fn=<NllLossBackward0>)
61000
6200 tensor(5.5360, device='cuda:0', grad_fn=<NllLossBackward0>)
62000
6300 tensor(5.3952, device='cuda:0', grad_fn=<NllLossBackward0>)
63000
6400 tensor(5.3132, device='cuda:0', grad_fn=<NllLossBackward0>)
64000
6500 tensor(5.1732, device='cuda:0', grad_fn=<NllLossBackward0>)
65000
6600 tensor(5.3505, device='cuda:0', grad_fn=<NllLossBackward0>)
66000
6700 tensor(5.2919, device='cuda:0', grad_fn=<NllLossBackward0>)
67000
6800 tensor(5.5064, device='cuda:0', grad_fn=<NllLossBackward0>)
68000
6900 tensor(5.4881, device='cuda:0', grad_fn=<NllLossBackward0>)
69000
7000 tensor(5.3978, device='cuda:0', grad_fn=<NllLossBackward0>)
70000
7100 tensor(5.2030, device='cuda:0', grad_fn=<NllLossBackward0>)
71000
7200 tensor(5.2738, device='cuda:0', grad_fn=<NllLossBackward0>)
72000
7300 tensor(5.5317, device='cuda:0', grad_fn=<NllLossBackward0>)
73000
7400 tensor(5.3487, device='cuda:0', grad_fn=<NllLossBackward0>)
74000
7500 tensor(5.2133, device='cuda:0', grad_fn=<NllLossBackward0>)
75000
7600 tensor(5.2878, device='cuda:0', grad_fn=<NllLossBackward0>)
76000
7700 tensor(5.3644, device='cuda:0', grad_fn=<NllLossBackward0>)
77000
7800 tensor(5.4711, device='cuda:0', grad_fn=<NllLossBackward0>)
78000
7900 tensor(5.1445, device='cuda:0', grad_fn=<NllLossBackward0>)
79000
8000 tensor(5.2138, device='cuda:0', grad_fn=<NllLossBackward0>)
80000
8100 tensor(5.3741, device='cuda:0', grad_fn=<NllLossBackward0>)
81000
8200 tensor(5.3893, device='cuda:0', grad_fn=<NllLossBackward0>)
82000
8300 tensor(5.3492, device='cuda:0', grad_fn=<NllLossBackward0>)
83000
8400 tensor(5.4797, device='cuda:0', grad_fn=<NllLossBackward0>)
84000
8500 tensor(5.4501, device='cuda:0', grad_fn=<NllLossBackward0>)
85000
8600 tensor(5.4600, device='cuda:0', grad_fn=<NllLossBackward0>)
86000
8700 tensor(5.5758, device='cuda:0', grad_fn=<NllLossBackward0>)
87000
8800 tensor(5.4493, device='cuda:0', grad_fn=<NllLossBackward0>)
88000
8900 tensor(5.3035, device='cuda:0', grad_fn=<NllLossBackward0>)
89000
9000 tensor(5.4164, device='cuda:0', grad_fn=<NllLossBackward0>)
90000
9100 tensor(5.4273, device='cuda:0', grad_fn=<NllLossBackward0>)
91000
9200 tensor(5.3343, device='cuda:0', grad_fn=<NllLossBackward0>)
92000
9300 tensor(5.3797, device='cuda:0', grad_fn=<NllLossBackward0>)
93000
9400 tensor(5.2260, device='cuda:0', grad_fn=<NllLossBackward0>)
94000
9500 tensor(5.3006, device='cuda:0', grad_fn=<NllLossBackward0>)
95000
9600 tensor(5.4211, device='cuda:0', grad_fn=<NllLossBackward0>)
96000
9700 tensor(5.3615, device='cuda:0', grad_fn=<NllLossBackward0>)
97000
9800 tensor(5.4089, device='cuda:0', grad_fn=<NllLossBackward0>)
9900 tensor(5.2200, device='cuda:0', grad_fn=<NllLossBackward0>)
98000
10000 tensor(5.4428, device='cuda:0', grad_fn=<NllLossBackward0>)
99000
10100 tensor(5.3219, device='cuda:0', grad_fn=<NllLossBackward0>)
100000
10200 tensor(5.2692, device='cuda:0', grad_fn=<NllLossBackward0>)
101000
10300 tensor(5.3854, device='cuda:0', grad_fn=<NllLossBackward0>)
102000
10400 tensor(5.3984, device='cuda:0', grad_fn=<NllLossBackward0>)
103000
10500 tensor(5.4516, device='cuda:0', grad_fn=<NllLossBackward0>)
104000
10600 tensor(5.5380, device='cuda:0', grad_fn=<NllLossBackward0>)
105000
10700 tensor(5.3724, device='cuda:0', grad_fn=<NllLossBackward0>)
106000
10800 tensor(5.3862, device='cuda:0', grad_fn=<NllLossBackward0>)
107000
10900 tensor(5.3102, device='cuda:0', grad_fn=<NllLossBackward0>)
108000
11000 tensor(5.5487, device='cuda:0', grad_fn=<NllLossBackward0>)
109000
11100 tensor(5.1684, device='cuda:0', grad_fn=<NllLossBackward0>)
110000
11200 tensor(5.3303, device='cuda:0', grad_fn=<NllLossBackward0>)
111000
11300 tensor(5.3537, device='cuda:0', grad_fn=<NllLossBackward0>)
112000
11400 tensor(5.3064, device='cuda:0', grad_fn=<NllLossBackward0>)
113000
11500 tensor(5.3775, device='cuda:0', grad_fn=<NllLossBackward0>)
114000
11600 tensor(5.3649, device='cuda:0', grad_fn=<NllLossBackward0>)
115000
11700 tensor(5.2920, device='cuda:0', grad_fn=<NllLossBackward0>)
116000
11800 tensor(5.4908, device='cuda:0', grad_fn=<NllLossBackward0>)
117000
11900 tensor(5.3293, device='cuda:0', grad_fn=<NllLossBackward0>)
118000
12000 tensor(5.3926, device='cuda:0', grad_fn=<NllLossBackward0>)
119000
12100 tensor(5.4045, device='cuda:0', grad_fn=<NllLossBackward0>)
120000
12200 tensor(5.4246, device='cuda:0', grad_fn=<NllLossBackward0>)
121000
12300 tensor(5.5096, device='cuda:0', grad_fn=<NllLossBackward0>)
122000
12400 tensor(5.3884, device='cuda:0', grad_fn=<NllLossBackward0>)
123000
12500 tensor(5.3057, device='cuda:0', grad_fn=<NllLossBackward0>)
124000
12600 tensor(5.3466, device='cuda:0', grad_fn=<NllLossBackward0>)
125000
12700 tensor(5.2898, device='cuda:0', grad_fn=<NllLossBackward0>)
126000
12800 tensor(5.4714, device='cuda:0', grad_fn=<NllLossBackward0>)
127000
12900 tensor(5.3255, device='cuda:0', grad_fn=<NllLossBackward0>)
128000
13000 tensor(5.1438, device='cuda:0', grad_fn=<NllLossBackward0>)
129000
13100 tensor(5.3498, device='cuda:0', grad_fn=<NllLossBackward0>)
130000
13200 tensor(5.3890, device='cuda:0', grad_fn=<NllLossBackward0>)
131000
13300 tensor(5.2710, device='cuda:0', grad_fn=<NllLossBackward0>)
132000
13400 tensor(5.3541, device='cuda:0', grad_fn=<NllLossBackward0>)
133000
13500 tensor(5.3156, device='cuda:0', grad_fn=<NllLossBackward0>)
134000
13600 tensor(5.3957, device='cuda:0', grad_fn=<NllLossBackward0>)
135000
13700 tensor(5.3548, device='cuda:0', grad_fn=<NllLossBackward0>)
136000
13800 tensor(5.3577, device='cuda:0', grad_fn=<NllLossBackward0>)
137000
13900 tensor(5.2122, device='cuda:0', grad_fn=<NllLossBackward0>)
138000
14000 tensor(5.4587, device='cuda:0', grad_fn=<NllLossBackward0>)
139000
14100 tensor(5.1704, device='cuda:0', grad_fn=<NllLossBackward0>)
140000
14200 tensor(5.4419, device='cuda:0', grad_fn=<NllLossBackward0>)
141000
14300 tensor(5.4142, device='cuda:0', grad_fn=<NllLossBackward0>)
142000
14400 tensor(5.3058, device='cuda:0', grad_fn=<NllLossBackward0>)
143000
14500 tensor(5.4082, device='cuda:0', grad_fn=<NllLossBackward0>)
144000
14600 tensor(5.4414, device='cuda:0', grad_fn=<NllLossBackward0>)
145000
14700 tensor(5.3177, device='cuda:0', grad_fn=<NllLossBackward0>)
146000
14800 tensor(5.7665, device='cuda:0', grad_fn=<NllLossBackward0>)
147000
14900 tensor(5.4171, device='cuda:0', grad_fn=<NllLossBackward0>)
148000
15000 tensor(5.2698, device='cuda:0', grad_fn=<NllLossBackward0>)
149000
15100 tensor(5.4915, device='cuda:0', grad_fn=<NllLossBackward0>)
150000
15200 tensor(5.3576, device='cuda:0', grad_fn=<NllLossBackward0>)
151000
15300 tensor(5.3567, device='cuda:0', grad_fn=<NllLossBackward0>)
152000
15400 tensor(5.2379, device='cuda:0', grad_fn=<NllLossBackward0>)
153000
15500 tensor(5.4092, device='cuda:0', grad_fn=<NllLossBackward0>)
154000
15600 tensor(5.4042, device='cuda:0', grad_fn=<NllLossBackward0>)
155000
15700 tensor(5.3017, device='cuda:0', grad_fn=<NllLossBackward0>)
156000
15800 tensor(5.2188, device='cuda:0', grad_fn=<NllLossBackward0>)
157000
15900 tensor(5.3497, device='cuda:0', grad_fn=<NllLossBackward0>)
158000
16000 tensor(5.1718, device='cuda:0', grad_fn=<NllLossBackward0>)
159000
16100 tensor(5.2145, device='cuda:0', grad_fn=<NllLossBackward0>)
160000
16200 tensor(5.5591, device='cuda:0', grad_fn=<NllLossBackward0>)
161000
16300 tensor(5.3864, device='cuda:0', grad_fn=<NllLossBackward0>)
162000
16400 tensor(5.3719, device='cuda:0', grad_fn=<NllLossBackward0>)
163000
16500 tensor(5.4842, device='cuda:0', grad_fn=<NllLossBackward0>)
164000
16600 tensor(5.3329, device='cuda:0', grad_fn=<NllLossBackward0>)
165000
16700 tensor(5.3130, device='cuda:0', grad_fn=<NllLossBackward0>)
166000
16800 tensor(5.3903, device='cuda:0', grad_fn=<NllLossBackward0>)
167000
16900 tensor(5.1551, device='cuda:0', grad_fn=<NllLossBackward0>)
168000
17000 tensor(5.4229, device='cuda:0', grad_fn=<NllLossBackward0>)
169000
17100 tensor(5.2686, device='cuda:0', grad_fn=<NllLossBackward0>)
170000
17200 tensor(5.4099, device='cuda:0', grad_fn=<NllLossBackward0>)
171000
17300 tensor(5.4198, device='cuda:0', grad_fn=<NllLossBackward0>)
172000
17400 tensor(5.2162, device='cuda:0', grad_fn=<NllLossBackward0>)
173000
17500 tensor(5.2385, device='cuda:0', grad_fn=<NllLossBackward0>)
174000
17600 tensor(5.2786, device='cuda:0', grad_fn=<NllLossBackward0>)
175000
17700 tensor(5.2576, device='cuda:0', grad_fn=<NllLossBackward0>)
176000
17800 tensor(5.3158, device='cuda:0', grad_fn=<NllLossBackward0>)
177000
17900 tensor(5.2105, device='cuda:0', grad_fn=<NllLossBackward0>)
178000
18000 tensor(5.4627, device='cuda:0', grad_fn=<NllLossBackward0>)
179000
18100 tensor(5.3966, device='cuda:0', grad_fn=<NllLossBackward0>)
180000
18200 tensor(5.3108, device='cuda:0', grad_fn=<NllLossBackward0>)
181000
18300 tensor(5.3148, device='cuda:0', grad_fn=<NllLossBackward0>)
182000
18400 tensor(5.3321, device='cuda:0', grad_fn=<NllLossBackward0>)
183000
18500 tensor(5.3171, device='cuda:0', grad_fn=<NllLossBackward0>)
184000
18600 tensor(5.5247, device='cuda:0', grad_fn=<NllLossBackward0>)
185000
18700 tensor(5.3469, device='cuda:0', grad_fn=<NllLossBackward0>)
186000
18800 tensor(5.5092, device='cuda:0', grad_fn=<NllLossBackward0>)
187000
18900 tensor(5.3902, device='cuda:0', grad_fn=<NllLossBackward0>)
188000
19000 tensor(5.3904, device='cuda:0', grad_fn=<NllLossBackward0>)
189000
19100 tensor(5.5019, device='cuda:0', grad_fn=<NllLossBackward0>)
190000
19200 tensor(5.3838, device='cuda:0', grad_fn=<NllLossBackward0>)
191000
19300 tensor(5.4674, device='cuda:0', grad_fn=<NllLossBackward0>)
192000
19400 tensor(5.3223, device='cuda:0', grad_fn=<NllLossBackward0>)
193000
19500 tensor(5.3235, device='cuda:0', grad_fn=<NllLossBackward0>)
194000
19600 tensor(5.3589, device='cuda:0', grad_fn=<NllLossBackward0>)
195000
19700 tensor(5.4063, device='cuda:0', grad_fn=<NllLossBackward0>)
196000
19800 tensor(5.2838, device='cuda:0', grad_fn=<NllLossBackward0>)
197000
19900 tensor(5.2807, device='cuda:0', grad_fn=<NllLossBackward0>)
198000
20000 tensor(5.3038, device='cuda:0', grad_fn=<NllLossBackward0>)
199000
20100 tensor(5.2397, device='cuda:0', grad_fn=<NllLossBackward0>)
200000
20200 tensor(5.1723, device='cuda:0', grad_fn=<NllLossBackward0>)
201000
20300 tensor(5.2827, device='cuda:0', grad_fn=<NllLossBackward0>)
20400 tensor(5.3245, device='cuda:0', grad_fn=<NllLossBackward0>)
202000
20500 tensor(5.3303, device='cuda:0', grad_fn=<NllLossBackward0>)
203000
20600 tensor(5.5211, device='cuda:0', grad_fn=<NllLossBackward0>)
204000
20700 tensor(5.3629, device='cuda:0', grad_fn=<NllLossBackward0>)
205000
20800 tensor(5.1882, device='cuda:0', grad_fn=<NllLossBackward0>)
206000
20900 tensor(5.4671, device='cuda:0', grad_fn=<NllLossBackward0>)
207000
21000 tensor(5.3110, device='cuda:0', grad_fn=<NllLossBackward0>)
208000
21100 tensor(5.3181, device='cuda:0', grad_fn=<NllLossBackward0>)
209000
21200 tensor(5.1968, device='cuda:0', grad_fn=<NllLossBackward0>)
210000
21300 tensor(5.3940, device='cuda:0', grad_fn=<NllLossBackward0>)
211000
21400 tensor(5.4308, device='cuda:0', grad_fn=<NllLossBackward0>)
212000
21500 tensor(5.2127, device='cuda:0', grad_fn=<NllLossBackward0>)
213000
21600 tensor(5.2003, device='cuda:0', grad_fn=<NllLossBackward0>)
214000
21700 tensor(5.1881, device='cuda:0', grad_fn=<NllLossBackward0>)
215000
21800 tensor(5.3180, device='cuda:0', grad_fn=<NllLossBackward0>)
216000
21900 tensor(5.3197, device='cuda:0', grad_fn=<NllLossBackward0>)
217000
22000 tensor(5.3005, device='cuda:0', grad_fn=<NllLossBackward0>)
218000
22100 tensor(5.1776, device='cuda:0', grad_fn=<NllLossBackward0>)
219000
22200 tensor(5.0509, device='cuda:0', grad_fn=<NllLossBackward0>)
220000
22300 tensor(5.4807, device='cuda:0', grad_fn=<NllLossBackward0>)
221000
22400 tensor(5.2040, device='cuda:0', grad_fn=<NllLossBackward0>)
222000
22500 tensor(5.2161, device='cuda:0', grad_fn=<NllLossBackward0>)
223000
22600 tensor(5.4083, device='cuda:0', grad_fn=<NllLossBackward0>)
224000
22700 tensor(5.1619, device='cuda:0', grad_fn=<NllLossBackward0>)
225000
22800 tensor(5.4301, device='cuda:0', grad_fn=<NllLossBackward0>)
226000
22900 tensor(5.4791, device='cuda:0', grad_fn=<NllLossBackward0>)
227000
23000 tensor(5.3785, device='cuda:0', grad_fn=<NllLossBackward0>)
228000
23100 tensor(5.3705, device='cuda:0', grad_fn=<NllLossBackward0>)
229000
23200 tensor(5.3633, device='cuda:0', grad_fn=<NllLossBackward0>)
230000
23300 tensor(5.4443, device='cuda:0', grad_fn=<NllLossBackward0>)
231000
23400 tensor(5.4496, device='cuda:0', grad_fn=<NllLossBackward0>)
232000
23500 tensor(5.2961, device='cuda:0', grad_fn=<NllLossBackward0>)
233000
23600 tensor(5.2603, device='cuda:0', grad_fn=<NllLossBackward0>)
234000
23700 tensor(5.2793, device='cuda:0', grad_fn=<NllLossBackward0>)
235000
23800 tensor(5.1461, device='cuda:0', grad_fn=<NllLossBackward0>)
236000
23900 tensor(5.2376, device='cuda:0', grad_fn=<NllLossBackward0>)
237000
24000 tensor(5.2269, device='cuda:0', grad_fn=<NllLossBackward0>)
238000
24100 tensor(5.3154, device='cuda:0', grad_fn=<NllLossBackward0>)
239000
24200 tensor(5.4852, device='cuda:0', grad_fn=<NllLossBackward0>)
240000
24300 tensor(5.3785, device='cuda:0', grad_fn=<NllLossBackward0>)
241000
24400 tensor(5.5053, device='cuda:0', grad_fn=<NllLossBackward0>)
242000
24500 tensor(5.2987, device='cuda:0', grad_fn=<NllLossBackward0>)
243000
24600 tensor(5.4275, device='cuda:0', grad_fn=<NllLossBackward0>)
244000
24700 tensor(5.3283, device='cuda:0', grad_fn=<NllLossBackward0>)
245000
24800 tensor(5.3707, device='cuda:0', grad_fn=<NllLossBackward0>)
246000
24900 tensor(5.4294, device='cuda:0', grad_fn=<NllLossBackward0>)
247000
25000 tensor(5.4479, device='cuda:0', grad_fn=<NllLossBackward0>)
248000
25100 tensor(5.3629, device='cuda:0', grad_fn=<NllLossBackward0>)
249000
25200 tensor(5.3849, device='cuda:0', grad_fn=<NllLossBackward0>)
250000
25300 tensor(5.4124, device='cuda:0', grad_fn=<NllLossBackward0>)
251000
25400 tensor(5.3932, device='cuda:0', grad_fn=<NllLossBackward0>)
252000
25500 tensor(5.2893, device='cuda:0', grad_fn=<NllLossBackward0>)
253000
25600 tensor(5.5512, device='cuda:0', grad_fn=<NllLossBackward0>)
254000
25700 tensor(5.3227, device='cuda:0', grad_fn=<NllLossBackward0>)
255000
25800 tensor(5.4217, device='cuda:0', grad_fn=<NllLossBackward0>)
256000
25900 tensor(5.3637, device='cuda:0', grad_fn=<NllLossBackward0>)
257000
26000 tensor(5.3632, device='cuda:0', grad_fn=<NllLossBackward0>)
258000
26100 tensor(5.2841, device='cuda:0', grad_fn=<NllLossBackward0>)
259000
26200 tensor(5.2107, device='cuda:0', grad_fn=<NllLossBackward0>)
260000
26300 tensor(5.4024, device='cuda:0', grad_fn=<NllLossBackward0>)
261000
26400 tensor(5.4410, device='cuda:0', grad_fn=<NllLossBackward0>)
262000
26500 tensor(5.1685, device='cuda:0', grad_fn=<NllLossBackward0>)
263000
26600 tensor(5.5023, device='cuda:0', grad_fn=<NllLossBackward0>)
264000
26700 tensor(5.3654, device='cuda:0', grad_fn=<NllLossBackward0>)
265000
26800 tensor(5.5407, device='cuda:0', grad_fn=<NllLossBackward0>)
266000
26900 tensor(5.3000, device='cuda:0', grad_fn=<NllLossBackward0>)
267000
27000 tensor(5.2141, device='cuda:0', grad_fn=<NllLossBackward0>)
268000
27100 tensor(5.2490, device='cuda:0', grad_fn=<NllLossBackward0>)
269000
27200 tensor(5.2850, device='cuda:0', grad_fn=<NllLossBackward0>)
270000
27300 tensor(5.4811, device='cuda:0', grad_fn=<NllLossBackward0>)
271000
27400 tensor(5.3561, device='cuda:0', grad_fn=<NllLossBackward0>)
272000
27500 tensor(5.2602, device='cuda:0', grad_fn=<NllLossBackward0>)
273000
27600 tensor(5.5429, device='cuda:0', grad_fn=<NllLossBackward0>)
274000
27700 tensor(5.3794, device='cuda:0', grad_fn=<NllLossBackward0>)
275000
27800 tensor(5.3792, device='cuda:0', grad_fn=<NllLossBackward0>)
276000
27900 tensor(5.4873, device='cuda:0', grad_fn=<NllLossBackward0>)
277000
28000 tensor(5.3454, device='cuda:0', grad_fn=<NllLossBackward0>)
278000
28100 tensor(5.3113, device='cuda:0', grad_fn=<NllLossBackward0>)
279000
28200 tensor(5.4785, device='cuda:0', grad_fn=<NllLossBackward0>)
280000
28300 tensor(5.1013, device='cuda:0', grad_fn=<NllLossBackward0>)
281000
28400 tensor(5.5403, device='cuda:0', grad_fn=<NllLossBackward0>)
282000
28500 tensor(5.3676, device='cuda:0', grad_fn=<NllLossBackward0>)
283000
28600 tensor(5.3108, device='cuda:0', grad_fn=<NllLossBackward0>)
284000
28700 tensor(5.4403, device='cuda:0', grad_fn=<NllLossBackward0>)
285000
28800 tensor(5.4926, device='cuda:0', grad_fn=<NllLossBackward0>)
286000
28900 tensor(5.3638, device='cuda:0', grad_fn=<NllLossBackward0>)
287000
29000 tensor(5.2819, device='cuda:0', grad_fn=<NllLossBackward0>)
288000
29100 tensor(5.0362, device='cuda:0', grad_fn=<NllLossBackward0>)
289000
29200 tensor(5.1871, device='cuda:0', grad_fn=<NllLossBackward0>)
290000
29300 tensor(5.4697, device='cuda:0', grad_fn=<NllLossBackward0>)
291000
29400 tensor(5.5909, device='cuda:0', grad_fn=<NllLossBackward0>)
292000
29500 tensor(5.3807, device='cuda:0', grad_fn=<NllLossBackward0>)
293000
29600 tensor(5.2398, device='cuda:0', grad_fn=<NllLossBackward0>)
294000
29700 tensor(5.3690, device='cuda:0', grad_fn=<NllLossBackward0>)
295000
29800 tensor(5.2220, device='cuda:0', grad_fn=<NllLossBackward0>)
296000
29900 tensor(5.4597, device='cuda:0', grad_fn=<NllLossBackward0>)
297000
30000 tensor(5.2205, device='cuda:0', grad_fn=<NllLossBackward0>)
298000
30100 tensor(5.3061, device='cuda:0', grad_fn=<NllLossBackward0>)
299000
30200 tensor(5.2432, device='cuda:0', grad_fn=<NllLossBackward0>)
300000
30300 tensor(5.3527, device='cuda:0', grad_fn=<NllLossBackward0>)
301000
30400 tensor(5.1823, device='cuda:0', grad_fn=<NllLossBackward0>)
302000
30500 tensor(5.3526, device='cuda:0', grad_fn=<NllLossBackward0>)
30600 tensor(5.3318, device='cuda:0', grad_fn=<NllLossBackward0>)
303000
30700 tensor(5.3634, device='cuda:0', grad_fn=<NllLossBackward0>)
304000
30800 tensor(5.3571, device='cuda:0', grad_fn=<NllLossBackward0>)
305000
30900 tensor(5.3875, device='cuda:0', grad_fn=<NllLossBackward0>)
306000
31000 tensor(5.4983, device='cuda:0', grad_fn=<NllLossBackward0>)
307000
31100 tensor(5.1554, device='cuda:0', grad_fn=<NllLossBackward0>)
308000
31200 tensor(5.1952, device='cuda:0', grad_fn=<NllLossBackward0>)
309000
31300 tensor(5.4546, device='cuda:0', grad_fn=<NllLossBackward0>)
310000
31400 tensor(5.2307, device='cuda:0', grad_fn=<NllLossBackward0>)
311000
31500 tensor(5.4188, device='cuda:0', grad_fn=<NllLossBackward0>)
312000
31600 tensor(5.4085, device='cuda:0', grad_fn=<NllLossBackward0>)
313000
31700 tensor(5.3744, device='cuda:0', grad_fn=<NllLossBackward0>)
314000
31800 tensor(5.4766, device='cuda:0', grad_fn=<NllLossBackward0>)
315000
31900 tensor(5.1062, device='cuda:0', grad_fn=<NllLossBackward0>)
316000
32000 tensor(5.2924, device='cuda:0', grad_fn=<NllLossBackward0>)
317000
32100 tensor(5.1728, device='cuda:0', grad_fn=<NllLossBackward0>)
318000
32200 tensor(5.4863, device='cuda:0', grad_fn=<NllLossBackward0>)
319000
32300 tensor(5.4748, device='cuda:0', grad_fn=<NllLossBackward0>)
320000
32400 tensor(5.4518, device='cuda:0', grad_fn=<NllLossBackward0>)
321000
32500 tensor(5.2752, device='cuda:0', grad_fn=<NllLossBackward0>)
322000
32600 tensor(5.3822, device='cuda:0', grad_fn=<NllLossBackward0>)
323000
32700 tensor(5.3088, device='cuda:0', grad_fn=<NllLossBackward0>)
324000
32800 tensor(5.5403, device='cuda:0', grad_fn=<NllLossBackward0>)
325000
32900 tensor(5.4000, device='cuda:0', grad_fn=<NllLossBackward0>)
326000
33000 tensor(5.1837, device='cuda:0', grad_fn=<NllLossBackward0>)
327000
33100 tensor(5.3888, device='cuda:0', grad_fn=<NllLossBackward0>)
328000
33200 tensor(5.4849, device='cuda:0', grad_fn=<NllLossBackward0>)
329000
33300 tensor(5.2471, device='cuda:0', grad_fn=<NllLossBackward0>)
330000
33400 tensor(5.5246, device='cuda:0', grad_fn=<NllLossBackward0>)
331000
33500 tensor(5.3479, device='cuda:0', grad_fn=<NllLossBackward0>)
332000
33600 tensor(5.3043, device='cuda:0', grad_fn=<NllLossBackward0>)
333000
33700 tensor(5.3487, device='cuda:0', grad_fn=<NllLossBackward0>)
334000
33800 tensor(5.4368, device='cuda:0', grad_fn=<NllLossBackward0>)
335000
33900 tensor(5.1620, device='cuda:0', grad_fn=<NllLossBackward0>)
336000
34000 tensor(5.3873, device='cuda:0', grad_fn=<NllLossBackward0>)
337000
34100 tensor(5.3545, device='cuda:0', grad_fn=<NllLossBackward0>)
338000
34200 tensor(5.4001, device='cuda:0', grad_fn=<NllLossBackward0>)
339000
34300 tensor(5.1902, device='cuda:0', grad_fn=<NllLossBackward0>)
340000
34400 tensor(5.6453, device='cuda:0', grad_fn=<NllLossBackward0>)
341000
34500 tensor(5.5124, device='cuda:0', grad_fn=<NllLossBackward0>)
342000
34600 tensor(5.4069, device='cuda:0', grad_fn=<NllLossBackward0>)
343000
34700 tensor(5.4734, device='cuda:0', grad_fn=<NllLossBackward0>)
344000
34800 tensor(5.5014, device='cuda:0', grad_fn=<NllLossBackward0>)
345000
34900 tensor(5.5412, device='cuda:0', grad_fn=<NllLossBackward0>)
346000
35000 tensor(5.5132, device='cuda:0', grad_fn=<NllLossBackward0>)
347000
35100 tensor(5.3455, device='cuda:0', grad_fn=<NllLossBackward0>)
348000
35200 tensor(5.2694, device='cuda:0', grad_fn=<NllLossBackward0>)
349000
35300 tensor(5.4988, device='cuda:0', grad_fn=<NllLossBackward0>)
350000
35400 tensor(5.1485, device='cuda:0', grad_fn=<NllLossBackward0>)
351000
35500 tensor(5.2299, device='cuda:0', grad_fn=<NllLossBackward0>)
352000
35600 tensor(5.3643, device='cuda:0', grad_fn=<NllLossBackward0>)
353000
35700 tensor(5.2247, device='cuda:0', grad_fn=<NllLossBackward0>)
354000
35800 tensor(5.3615, device='cuda:0', grad_fn=<NllLossBackward0>)
355000
35900 tensor(5.3453, device='cuda:0', grad_fn=<NllLossBackward0>)
356000
36000 tensor(5.1217, device='cuda:0', grad_fn=<NllLossBackward0>)
357000
36100 tensor(5.4909, device='cuda:0', grad_fn=<NllLossBackward0>)
358000
36200 tensor(5.3382, device='cuda:0', grad_fn=<NllLossBackward0>)
359000
36300 tensor(5.6225, device='cuda:0', grad_fn=<NllLossBackward0>)
360000
36400 tensor(5.3167, device='cuda:0', grad_fn=<NllLossBackward0>)
361000
36500 tensor(5.3458, device='cuda:0', grad_fn=<NllLossBackward0>)
362000
36600 tensor(5.3608, device='cuda:0', grad_fn=<NllLossBackward0>)
363000
36700 tensor(5.1660, device='cuda:0', grad_fn=<NllLossBackward0>)
364000
36800 tensor(5.2737, device='cuda:0', grad_fn=<NllLossBackward0>)
365000
36900 tensor(5.3883, device='cuda:0', grad_fn=<NllLossBackward0>)
366000
37000 tensor(5.2783, device='cuda:0', grad_fn=<NllLossBackward0>)
367000
37100 tensor(5.3110, device='cuda:0', grad_fn=<NllLossBackward0>)
368000
37200 tensor(5.3794, device='cuda:0', grad_fn=<NllLossBackward0>)
369000
37300 tensor(5.2802, device='cuda:0', grad_fn=<NllLossBackward0>)
370000
37400 tensor(5.6133, device='cuda:0', grad_fn=<NllLossBackward0>)
371000
37500 tensor(5.3138, device='cuda:0', grad_fn=<NllLossBackward0>)
372000
37600 tensor(5.3083, device='cuda:0', grad_fn=<NllLossBackward0>)
373000
37700 tensor(5.4860, device='cuda:0', grad_fn=<NllLossBackward0>)
374000
37800 tensor(5.3216, device='cuda:0', grad_fn=<NllLossBackward0>)
375000
37900 tensor(5.2969, device='cuda:0', grad_fn=<NllLossBackward0>)
376000
38000 tensor(5.3759, device='cuda:0', grad_fn=<NllLossBackward0>)
377000
38100 tensor(5.3914, device='cuda:0', grad_fn=<NllLossBackward0>)
378000
38200 tensor(5.4089, device='cuda:0', grad_fn=<NllLossBackward0>)
379000
38300 tensor(5.3068, device='cuda:0', grad_fn=<NllLossBackward0>)
380000
38400 tensor(5.3798, device='cuda:0', grad_fn=<NllLossBackward0>)
381000
38500 tensor(5.4051, device='cuda:0', grad_fn=<NllLossBackward0>)
382000
38600 tensor(5.3471, device='cuda:0', grad_fn=<NllLossBackward0>)
383000
38700 tensor(5.3415, device='cuda:0', grad_fn=<NllLossBackward0>)
384000
38800 tensor(5.4310, device='cuda:0', grad_fn=<NllLossBackward0>)
385000
38900 tensor(5.5029, device='cuda:0', grad_fn=<NllLossBackward0>)
386000
39000 tensor(5.2021, device='cuda:0', grad_fn=<NllLossBackward0>)
387000
39100 tensor(5.4283, device='cuda:0', grad_fn=<NllLossBackward0>)
388000
39200 tensor(5.5158, device='cuda:0', grad_fn=<NllLossBackward0>)
389000
39300 tensor(5.3452, device='cuda:0', grad_fn=<NllLossBackward0>)
390000
39400 tensor(5.4111, device='cuda:0', grad_fn=<NllLossBackward0>)
391000
39500 tensor(5.4969, device='cuda:0', grad_fn=<NllLossBackward0>)
392000
39600 tensor(5.1952, device='cuda:0', grad_fn=<NllLossBackward0>)
393000
39700 tensor(5.1946, device='cuda:0', grad_fn=<NllLossBackward0>)
394000
39800 tensor(5.3234, device='cuda:0', grad_fn=<NllLossBackward0>)
395000
39900 tensor(5.1354, device='cuda:0', grad_fn=<NllLossBackward0>)
396000
40000 tensor(5.2210, device='cuda:0', grad_fn=<NllLossBackward0>)
397000
40100 tensor(5.3133, device='cuda:0', grad_fn=<NllLossBackward0>)
398000
40200 tensor(5.2990, device='cuda:0', grad_fn=<NllLossBackward0>)
399000
40300 tensor(5.3684, device='cuda:0', grad_fn=<NllLossBackward0>)
40400 tensor(5.3700, device='cuda:0', grad_fn=<NllLossBackward0>)
400000
40500 tensor(5.2911, device='cuda:0', grad_fn=<NllLossBackward0>)
401000
40600 tensor(5.3497, device='cuda:0', grad_fn=<NllLossBackward0>)
402000
40700 tensor(5.3981, device='cuda:0', grad_fn=<NllLossBackward0>)
403000
40800 tensor(5.3436, device='cuda:0', grad_fn=<NllLossBackward0>)
404000
40900 tensor(5.2978, device='cuda:0', grad_fn=<NllLossBackward0>)
405000
41000 tensor(5.3420, device='cuda:0', grad_fn=<NllLossBackward0>)
406000
41100 tensor(5.3342, device='cuda:0', grad_fn=<NllLossBackward0>)
407000
41200 tensor(5.2226, device='cuda:0', grad_fn=<NllLossBackward0>)
408000
41300 tensor(5.3573, device='cuda:0', grad_fn=<NllLossBackward0>)
409000
41400 tensor(5.2448, device='cuda:0', grad_fn=<NllLossBackward0>)
410000
41500 tensor(5.3863, device='cuda:0', grad_fn=<NllLossBackward0>)
411000
41600 tensor(5.3051, device='cuda:0', grad_fn=<NllLossBackward0>)
412000
41700 tensor(5.3294, device='cuda:0', grad_fn=<NllLossBackward0>)
413000
41800 tensor(5.3191, device='cuda:0', grad_fn=<NllLossBackward0>)
414000
41900 tensor(5.3289, device='cuda:0', grad_fn=<NllLossBackward0>)
415000
42000 tensor(5.4860, device='cuda:0', grad_fn=<NllLossBackward0>)
416000
42100 tensor(5.2358, device='cuda:0', grad_fn=<NllLossBackward0>)
417000
42200 tensor(5.3253, device='cuda:0', grad_fn=<NllLossBackward0>)
418000
42300 tensor(5.3869, device='cuda:0', grad_fn=<NllLossBackward0>)
419000
42400 tensor(5.2062, device='cuda:0', grad_fn=<NllLossBackward0>)
420000
42500 tensor(5.3712, device='cuda:0', grad_fn=<NllLossBackward0>)
421000
42600 tensor(5.1718, device='cuda:0', grad_fn=<NllLossBackward0>)
422000
42700 tensor(5.4735, device='cuda:0', grad_fn=<NllLossBackward0>)
423000
42800 tensor(5.3973, device='cuda:0', grad_fn=<NllLossBackward0>)
424000
42900 tensor(5.2447, device='cuda:0', grad_fn=<NllLossBackward0>)
425000
43000 tensor(5.3896, device='cuda:0', grad_fn=<NllLossBackward0>)
426000
43100 tensor(5.3916, device='cuda:0', grad_fn=<NllLossBackward0>)
427000
43200 tensor(5.2044, device='cuda:0', grad_fn=<NllLossBackward0>)
428000
43300 tensor(5.2167, device='cuda:0', grad_fn=<NllLossBackward0>)
429000
43400 tensor(5.3933, device='cuda:0', grad_fn=<NllLossBackward0>)
430000
43500 tensor(5.1078, device='cuda:0', grad_fn=<NllLossBackward0>)
431000
43600 tensor(5.3045, device='cuda:0', grad_fn=<NllLossBackward0>)
432000

device = 'cuda'
torch.cuda.empty_cache()
model = SimpleBigramNeuralLanguageModel(vocab_size, embed_size).to(device)
model.load_state_dict(torch.load(f'model-bigram_final.bin'))
model.eval()

ixs = torch.tensor(vocab.forward(['will'])).to(device)

out = model(ixs)
top = torch.topk(out[0], 10)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))

[('be', 11, 0.2570849657058716),
 ('<unk>', 0, 0.07411641627550125),
 ('not', 22, 0.05940083786845207),
 ('have', 28, 0.02751326560974121),
 ('bo', 167, 0.014936885796487331),
 ('make', 116, 0.013943656347692013),
 ('give', 193, 0.011286991648375988),
 ('take', 153, 0.011171611957252026),
 ('do', 86, 0.010088067501783371),
 ('he', 20, 0.009703895077109337)]

vocab = train_dataset.vocab
ixs = torch.tensor(vocab.forward(['cerned.'])).to(device)

out = model(ixs)
top = torch.topk(out[0], 10)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))

[('<unk>', 0, 0.19996878504753113),
 ('and', 3, 0.05288130044937134),
 ('of', 2, 0.042051784694194794),
 ('the', 1, 0.026572922244668007),
 ('to', 4, 0.022689413279294968),
 ('in', 6, 0.015904497355222702),
 ('The', 17, 0.012827681377530098),
 ('a', 5, 0.00961760152131319),
 ('for', 8, 0.008938422426581383),
 ('</s>', 32, 0.00840282253921032)]

cos = nn.CosineSimilarity(dim=1, eps=1e-6)

embeddings = model.model[0].weight

vec = embeddings[vocab['cerned.']]

similarities = cos(vec, embeddings)

top = torch.topk(similarities, 10)

top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))

[('<unk>', 0, 1.0),
 ('particular,', 14538, 0.24527804553508759),
 ('revolution.', 20446, 0.23776617646217346),
 ('Territory.', 14189, 0.23417341709136963),
 ('or-', 2261, 0.22888363897800446),
 ('3', 479, 0.2288265973329544),
 ('speak.', 13722, 0.2252315878868103),
 ('attend.', 19397, 0.22110989689826965),
 ('say,', 1455, 0.22106117010116577),
 ('Lee.', 15326, 0.21764159202575684)]

def get_values_from_model(presc_word, model, vocab, k):
    ixs =  torch.tensor(vocab.forward([presc_word])).to(device)
    out = model(ixs)
    top = torch.topk(out[0], k)
    top_indices = top.indices.tolist()
    top_probs = top.values.tolist()
    top_words = vocab.lookup_tokens(top_indices)
    return list(zip(top_words, top_probs))

def gonito_format(dic):
    tab = summarize_probs_unk(dic)
    result = ''
    for element in tab[:-1]:
        result+=str(element[0])+':'+str(element[1])+'\t'
    result+=':'+ str(tab[-1][1])+'\n'
    return result

def summarize_probs_unk(dic):
    if '<unk>' in dic.keys():
        probsum = sum(float(val) for key, val in dic.items())
        for key in dic:
            dic[key] = dic[key]/probsum ###leave some space for wildcard
        wildcard = dic['<unk>']
        del dic['<unk>']
        tab = [(key, val) for key, val in dic.items()]
        tab.append(('<unk>', wildcard))
    else:
        probsum = sum(float(val) for key, val in dic.items())
        for key in dic:
            dic[key] = dic[key]/(probsum*(1+wildcard_minweight)) #plus, becouse it's denominator
        tab = [(key, val) for key, val in dic.items()]
        tab.append(('<unk>', 1-sum([val for val in dic.values()])))
    return tab

model.load_state_dict(torch.load('model-bigram_final.bin'))

<All keys matched successfully>


with lzma.open(test_file, 'rt') as file:
    predict_words = []
    results = []
    for line in file:
#         print(line)
        line = preprocess(line) #get only relevant
        split = line.split('\t')
        predict_words.append(get_last_word(split[0])) #get_first_word(split[1])
    vocab = train_dataset.vocab
    for presc_word in predict_words:
        results.append(dict(get_values_from_model(presc_word, model, vocab, k=k)))
    with open(out_file, 'w') as outfile:
        for elem in results:
            outfile.write(gonito_format(elem))

261 KiB Raw Permalink Blame History Unescape Escape

Definicja sieci

261 KiB

Raw Permalink Blame History