123 KiB
123 KiB
import itertools
import lzma
import numpy as np
import regex as re
import torch
from torch import nn
from torch.utils.data import IterableDataset, DataLoader
from torchtext.vocab import build_vocab_from_iteratorvocab_size = 10000
embed_size = 200
hidden_size = 100
batch_size = 1000
device = 'cuda'
path_to_train = 'train/in.tsv.xz'
path_to_model = 'model1.bin'
def clean_line(line: str):
# Preprocessing
separated = line.split('\t')
prefix = separated[6].replace(r'\n', ' ')
suffix = separated[7].replace(r'\n', ' ')
return prefix + ' ' + suffix
def get_words_from_line(line):
line = clean_line(line)
for word in line.split():
yield word
def get_word_lines_from_file(file_name):
with lzma.open(file_name, mode='rt', encoding='utf-8') as fid:
for line in fid:
yield get_words_from_line(line)
def double_look_ahead_iterator(gen):
prev_prev = None
prev = None
for item in gen:
if prev_prev is not None:
yield np.asarray((prev_prev, prev, item))
prev_prev = prev
prev = item
def prediction(words, model, top) -> str:
words_tensor = [train_dataset.vocab.forward([word]) for word in words]
ixs = torch.tensor(words_tensor).view(-1).to(device)
out = model(ixs)
top = torch.topk(out[0], top)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
zipped = list(zip(top_words, top_probs))
for index, element in enumerate(zipped):
unk = None
if '<unk>' in element:
unk = zipped.pop(index)
zipped.append(('', unk[1]))
break
if unk is None:
zipped[-1] = ('', zipped[-1][1])
return ' '.join([f'{x[0]}:{x[1]:.5f}' for x in zipped])
def create_outputs(folder_name, model, top):
print(f'Creating outputs in {folder_name}')
with lzma.open(f'{folder_name}/in.tsv.xz', mode='rt', encoding='utf-8') as fid:
with open(f'{folder_name}/out-top={top}.tsv', 'w', encoding='utf-8', newline='\n') as f:
for line in fid:
separated = line.split('\t')
prefix = separated[6].replace(r'\n', ' ').split()[-2:]
output_line = prediction(prefix, model, top)
f.write(output_line + '\n')
def train_model(lr):
model = SimpleTrigramNeuralLanguageModel(vocab_size, embed_size, hidden_size).to(device)
data = DataLoader(train_dataset, batch_size=batch_size)
optimizer = torch.optim.Adam(model.parameters(), lr=lr)
criterion = torch.nn.NLLLoss()
model.train()
step = 0
for batch in data:
x = batch[:, :2]
y = batch[:, 2]
x = x.to(device)
y = y.to(device)
optimizer.zero_grad()
ypredicted = model(x)
loss = criterion(torch.log(ypredicted), y)
if step % 100 == 0:
print(step, loss)
step += 1
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), 10)
optimizer.step()
torch.save(model.state_dict(), path_to_model)
def with_hyperparams():
train_model(lr=0.0001)
model = SimpleTrigramNeuralLanguageModel(vocab_size, embed_size, hidden_size).to(device)
model.load_state_dict(torch.load(path_to_model))
model.eval()
for top in [200, 400, 600]:
create_outputs('dev-0', model, top)
create_outputs('test-A', model, top)
class Trigrams(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 double_look_ahead_iterator(
(self.vocab[t] for t in itertools.chain.from_iterable(get_word_lines_from_file(self.text_file))))
class SimpleTrigramNeuralLanguageModel(nn.Module):
def __init__(self, vocabulary_size, embedding_size, hidden_size):
super(SimpleTrigramNeuralLanguageModel, self).__init__()
self.embedding_size = embedding_size
self.embedding = nn.Embedding(vocabulary_size, embedding_size)
self.lin1 = nn.Linear(2 * embedding_size, hidden_size)
self.rel = nn.ReLU()
self.lin2 = nn.Linear(hidden_size, vocabulary_size)
self.sm = nn.Softmax()
def forward(self, x):
x = self.embedding(x).view((-1, 2 * self.embedding_size))
x = self.lin1(x)
x = self.rel(x)
x = self.lin2(x)
return self.sm(x)from google.colab import drive
drive.mount('/content/drive',force_remount=True)
%cd /content/drive/MyDrive/modelowanie_jezykaMounted at /content/drive /content/drive/MyDrive/modelowanie_jezyka
vocab = build_vocab_from_iterator(
get_word_lines_from_file(path_to_train),
max_tokens=vocab_size,
specials=['<unk>']
)
vocab.set_default_index(vocab['<unk>'])
train_dataset = Trigrams(path_to_train, vocab_size)with_hyperparams()<ipython-input-7-15bd6173eff4>:131: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument. return self.sm(x)
0 tensor(9.2570, device='cuda:0', grad_fn=<NllLossBackward0>) 100 tensor(8.3136, device='cuda:0', grad_fn=<NllLossBackward0>) 200 tensor(7.4358, device='cuda:0', grad_fn=<NllLossBackward0>) 300 tensor(7.2475, device='cuda:0', grad_fn=<NllLossBackward0>) 400 tensor(6.7645, device='cuda:0', grad_fn=<NllLossBackward0>) 500 tensor(6.5050, device='cuda:0', grad_fn=<NllLossBackward0>) 600 tensor(6.4014, device='cuda:0', grad_fn=<NllLossBackward0>) 700 tensor(6.6077, device='cuda:0', grad_fn=<NllLossBackward0>) 800 tensor(6.3927, device='cuda:0', grad_fn=<NllLossBackward0>) 900 tensor(6.0547, device='cuda:0', grad_fn=<NllLossBackward0>) 1000 tensor(6.1434, device='cuda:0', grad_fn=<NllLossBackward0>) 1100 tensor(5.8979, device='cuda:0', grad_fn=<NllLossBackward0>) 1200 tensor(6.1095, device='cuda:0', grad_fn=<NllLossBackward0>) 1300 tensor(6.1998, device='cuda:0', grad_fn=<NllLossBackward0>) 1400 tensor(5.9146, device='cuda:0', grad_fn=<NllLossBackward0>) 1500 tensor(5.7813, device='cuda:0', grad_fn=<NllLossBackward0>) 1600 tensor(5.9985, device='cuda:0', grad_fn=<NllLossBackward0>) 1700 tensor(5.3452, device='cuda:0', grad_fn=<NllLossBackward0>) 1800 tensor(5.7940, device='cuda:0', grad_fn=<NllLossBackward0>) 1900 tensor(5.5534, device='cuda:0', grad_fn=<NllLossBackward0>) 2000 tensor(5.6145, device='cuda:0', grad_fn=<NllLossBackward0>) 2100 tensor(5.8359, device='cuda:0', grad_fn=<NllLossBackward0>) 2200 tensor(5.7575, device='cuda:0', grad_fn=<NllLossBackward0>) 2300 tensor(5.8183, device='cuda:0', grad_fn=<NllLossBackward0>) 2400 tensor(5.7377, device='cuda:0', grad_fn=<NllLossBackward0>) 2500 tensor(5.8494, device='cuda:0', grad_fn=<NllLossBackward0>) 2600 tensor(5.8213, device='cuda:0', grad_fn=<NllLossBackward0>) 2700 tensor(5.6311, device='cuda:0', grad_fn=<NllLossBackward0>) 2800 tensor(5.5126, device='cuda:0', grad_fn=<NllLossBackward0>) 2900 tensor(5.4556, device='cuda:0', grad_fn=<NllLossBackward0>) 3000 tensor(5.2281, device='cuda:0', grad_fn=<NllLossBackward0>) 3100 tensor(5.4690, device='cuda:0', grad_fn=<NllLossBackward0>) 3200 tensor(5.2455, device='cuda:0', grad_fn=<NllLossBackward0>) 3300 tensor(5.8062, device='cuda:0', grad_fn=<NllLossBackward0>) 3400 tensor(5.8635, device='cuda:0', grad_fn=<NllLossBackward0>) 3500 tensor(5.5261, device='cuda:0', grad_fn=<NllLossBackward0>) 3600 tensor(5.4733, device='cuda:0', grad_fn=<NllLossBackward0>) 3700 tensor(5.3929, device='cuda:0', grad_fn=<NllLossBackward0>) 3800 tensor(5.5937, device='cuda:0', grad_fn=<NllLossBackward0>) 3900 tensor(5.4884, device='cuda:0', grad_fn=<NllLossBackward0>) 4000 tensor(5.1098, device='cuda:0', grad_fn=<NllLossBackward0>) 4100 tensor(5.4178, device='cuda:0', grad_fn=<NllLossBackward0>) 4200 tensor(5.3855, device='cuda:0', grad_fn=<NllLossBackward0>) 4300 tensor(5.4540, device='cuda:0', grad_fn=<NllLossBackward0>) 4400 tensor(5.2868, device='cuda:0', grad_fn=<NllLossBackward0>) 4500 tensor(5.6309, device='cuda:0', grad_fn=<NllLossBackward0>) 4600 tensor(5.6222, device='cuda:0', grad_fn=<NllLossBackward0>) 4700 tensor(5.0256, device='cuda:0', grad_fn=<NllLossBackward0>) 4800 tensor(5.5779, device='cuda:0', grad_fn=<NllLossBackward0>) 4900 tensor(5.4185, device='cuda:0', grad_fn=<NllLossBackward0>) 5000 tensor(5.0892, device='cuda:0', grad_fn=<NllLossBackward0>) 5100 tensor(5.2962, device='cuda:0', grad_fn=<NllLossBackward0>) 5200 tensor(5.6080, device='cuda:0', grad_fn=<NllLossBackward0>) 5300 tensor(5.5294, device='cuda:0', grad_fn=<NllLossBackward0>) 5400 tensor(5.4334, device='cuda:0', grad_fn=<NllLossBackward0>) 5500 tensor(5.3486, device='cuda:0', grad_fn=<NllLossBackward0>) 5600 tensor(5.5954, device='cuda:0', grad_fn=<NllLossBackward0>) 5700 tensor(5.6030, device='cuda:0', grad_fn=<NllLossBackward0>) 5800 tensor(5.3286, device='cuda:0', grad_fn=<NllLossBackward0>) 5900 tensor(5.2896, device='cuda:0', grad_fn=<NllLossBackward0>) 6000 tensor(5.7665, device='cuda:0', grad_fn=<NllLossBackward0>) 6100 tensor(5.3989, device='cuda:0', grad_fn=<NllLossBackward0>) 6200 tensor(5.1637, device='cuda:0', grad_fn=<NllLossBackward0>) 6300 tensor(5.7359, device='cuda:0', grad_fn=<NllLossBackward0>) 6400 tensor(5.4357, device='cuda:0', grad_fn=<NllLossBackward0>) 6500 tensor(5.2306, device='cuda:0', grad_fn=<NllLossBackward0>) 6600 tensor(5.4952, device='cuda:0', grad_fn=<NllLossBackward0>) 6700 tensor(5.2873, device='cuda:0', grad_fn=<NllLossBackward0>) 6800 tensor(5.6516, device='cuda:0', grad_fn=<NllLossBackward0>) 6900 tensor(5.3410, device='cuda:0', grad_fn=<NllLossBackward0>) 7000 tensor(5.5064, device='cuda:0', grad_fn=<NllLossBackward0>) 7100 tensor(5.3542, device='cuda:0', grad_fn=<NllLossBackward0>) 7200 tensor(5.4427, device='cuda:0', grad_fn=<NllLossBackward0>) 7300 tensor(5.4514, device='cuda:0', grad_fn=<NllLossBackward0>) 7400 tensor(5.3294, device='cuda:0', grad_fn=<NllLossBackward0>) 7500 tensor(5.7200, device='cuda:0', grad_fn=<NllLossBackward0>) 7600 tensor(5.2620, device='cuda:0', grad_fn=<NllLossBackward0>) 7700 tensor(5.3428, device='cuda:0', grad_fn=<NllLossBackward0>) 7800 tensor(5.5721, device='cuda:0', grad_fn=<NllLossBackward0>) 7900 tensor(5.4102, device='cuda:0', grad_fn=<NllLossBackward0>) 8000 tensor(5.4482, device='cuda:0', grad_fn=<NllLossBackward0>) 8100 tensor(5.4963, device='cuda:0', grad_fn=<NllLossBackward0>) 8200 tensor(5.5347, device='cuda:0', grad_fn=<NllLossBackward0>) 8300 tensor(5.5172, device='cuda:0', grad_fn=<NllLossBackward0>) 8400 tensor(5.0658, device='cuda:0', grad_fn=<NllLossBackward0>) 8500 tensor(5.4056, device='cuda:0', grad_fn=<NllLossBackward0>) 8600 tensor(5.1450, device='cuda:0', grad_fn=<NllLossBackward0>) 8700 tensor(5.2514, device='cuda:0', grad_fn=<NllLossBackward0>) 8800 tensor(5.0214, device='cuda:0', grad_fn=<NllLossBackward0>) 8900 tensor(5.3136, device='cuda:0', grad_fn=<NllLossBackward0>) 9000 tensor(5.1997, device='cuda:0', grad_fn=<NllLossBackward0>) 9100 tensor(4.8348, device='cuda:0', grad_fn=<NllLossBackward0>) 9200 tensor(5.0372, device='cuda:0', grad_fn=<NllLossBackward0>) 9300 tensor(5.2466, device='cuda:0', grad_fn=<NllLossBackward0>) 9400 tensor(5.3362, device='cuda:0', grad_fn=<NllLossBackward0>) 9500 tensor(5.2651, device='cuda:0', grad_fn=<NllLossBackward0>) 9600 tensor(5.5185, device='cuda:0', grad_fn=<NllLossBackward0>) 9700 tensor(5.5460, device='cuda:0', grad_fn=<NllLossBackward0>) 9800 tensor(5.1778, device='cuda:0', grad_fn=<NllLossBackward0>) 9900 tensor(5.0272, device='cuda:0', grad_fn=<NllLossBackward0>) 10000 tensor(5.4087, device='cuda:0', grad_fn=<NllLossBackward0>) 10100 tensor(5.1365, device='cuda:0', grad_fn=<NllLossBackward0>) 10200 tensor(5.1908, device='cuda:0', grad_fn=<NllLossBackward0>) 10300 tensor(5.1816, device='cuda:0', grad_fn=<NllLossBackward0>) 10400 tensor(5.5366, device='cuda:0', grad_fn=<NllLossBackward0>) 10500 tensor(5.2595, device='cuda:0', grad_fn=<NllLossBackward0>) 10600 tensor(5.3326, device='cuda:0', grad_fn=<NllLossBackward0>) 10700 tensor(5.3983, device='cuda:0', grad_fn=<NllLossBackward0>) 10800 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>) 10900 tensor(5.3010, device='cuda:0', grad_fn=<NllLossBackward0>) 11000 tensor(5.2640, device='cuda:0', grad_fn=<NllLossBackward0>) 11100 tensor(4.9914, device='cuda:0', grad_fn=<NllLossBackward0>) 11200 tensor(5.2707, device='cuda:0', grad_fn=<NllLossBackward0>) 11300 tensor(5.1615, device='cuda:0', grad_fn=<NllLossBackward0>) 11400 tensor(5.3967, device='cuda:0', grad_fn=<NllLossBackward0>) 11500 tensor(5.0548, device='cuda:0', grad_fn=<NllLossBackward0>) 11600 tensor(5.2904, device='cuda:0', grad_fn=<NllLossBackward0>) 11700 tensor(5.0728, device='cuda:0', grad_fn=<NllLossBackward0>) 11800 tensor(5.3685, device='cuda:0', grad_fn=<NllLossBackward0>) 11900 tensor(5.2579, device='cuda:0', grad_fn=<NllLossBackward0>) 12000 tensor(5.4116, device='cuda:0', grad_fn=<NllLossBackward0>) 12100 tensor(4.9850, device='cuda:0', grad_fn=<NllLossBackward0>) 12200 tensor(5.1288, device='cuda:0', grad_fn=<NllLossBackward0>) 12300 tensor(5.5204, device='cuda:0', grad_fn=<NllLossBackward0>) 12400 tensor(5.2945, device='cuda:0', grad_fn=<NllLossBackward0>) 12500 tensor(5.2505, device='cuda:0', grad_fn=<NllLossBackward0>) 12600 tensor(4.8808, device='cuda:0', grad_fn=<NllLossBackward0>) 12700 tensor(4.6960, device='cuda:0', grad_fn=<NllLossBackward0>) 12800 tensor(5.2775, device='cuda:0', grad_fn=<NllLossBackward0>) 12900 tensor(5.4755, device='cuda:0', grad_fn=<NllLossBackward0>) 13000 tensor(4.8998, device='cuda:0', grad_fn=<NllLossBackward0>) 13100 tensor(5.0161, device='cuda:0', grad_fn=<NllLossBackward0>) 13200 tensor(5.3238, device='cuda:0', grad_fn=<NllLossBackward0>) 13300 tensor(5.3577, device='cuda:0', grad_fn=<NllLossBackward0>) 13400 tensor(5.3376, device='cuda:0', grad_fn=<NllLossBackward0>) 13500 tensor(5.4877, device='cuda:0', grad_fn=<NllLossBackward0>) 13600 tensor(4.8643, device='cuda:0', grad_fn=<NllLossBackward0>) 13700 tensor(4.9484, device='cuda:0', grad_fn=<NllLossBackward0>) 13800 tensor(5.1129, device='cuda:0', grad_fn=<NllLossBackward0>) 13900 tensor(5.3575, device='cuda:0', grad_fn=<NllLossBackward0>) 14000 tensor(5.4240, device='cuda:0', grad_fn=<NllLossBackward0>) 14100 tensor(5.1704, device='cuda:0', grad_fn=<NllLossBackward0>) 14200 tensor(5.1959, device='cuda:0', grad_fn=<NllLossBackward0>) 14300 tensor(5.3272, device='cuda:0', grad_fn=<NllLossBackward0>) 14400 tensor(5.4148, device='cuda:0', grad_fn=<NllLossBackward0>) 14500 tensor(5.3436, device='cuda:0', grad_fn=<NllLossBackward0>) 14600 tensor(5.2994, device='cuda:0', grad_fn=<NllLossBackward0>) 14700 tensor(5.3515, device='cuda:0', grad_fn=<NllLossBackward0>) 14800 tensor(5.0899, device='cuda:0', grad_fn=<NllLossBackward0>) 14900 tensor(5.0274, device='cuda:0', grad_fn=<NllLossBackward0>) 15000 tensor(5.2740, device='cuda:0', grad_fn=<NllLossBackward0>) 15100 tensor(5.4064, device='cuda:0', grad_fn=<NllLossBackward0>) 15200 tensor(5.1399, device='cuda:0', grad_fn=<NllLossBackward0>) 15300 tensor(5.0051, device='cuda:0', grad_fn=<NllLossBackward0>) 15400 tensor(5.0513, device='cuda:0', grad_fn=<NllLossBackward0>) 15500 tensor(5.2234, device='cuda:0', grad_fn=<NllLossBackward0>) 15600 tensor(5.1812, device='cuda:0', grad_fn=<NllLossBackward0>) 15700 tensor(5.2933, device='cuda:0', grad_fn=<NllLossBackward0>) 15800 tensor(5.3839, device='cuda:0', grad_fn=<NllLossBackward0>) 15900 tensor(5.3919, device='cuda:0', grad_fn=<NllLossBackward0>) 16000 tensor(5.0378, device='cuda:0', grad_fn=<NllLossBackward0>) 16100 tensor(5.2975, device='cuda:0', grad_fn=<NllLossBackward0>) 16200 tensor(5.2050, device='cuda:0', grad_fn=<NllLossBackward0>) 16300 tensor(4.9146, device='cuda:0', grad_fn=<NllLossBackward0>) 16400 tensor(5.0289, device='cuda:0', grad_fn=<NllLossBackward0>) 16500 tensor(5.5424, device='cuda:0', grad_fn=<NllLossBackward0>) 16600 tensor(4.9872, device='cuda:0', grad_fn=<NllLossBackward0>) 16700 tensor(5.3475, device='cuda:0', grad_fn=<NllLossBackward0>) 16800 tensor(5.1461, device='cuda:0', grad_fn=<NllLossBackward0>) 16900 tensor(5.4058, device='cuda:0', grad_fn=<NllLossBackward0>) 17000 tensor(4.3672, device='cuda:0', grad_fn=<NllLossBackward0>) 17100 tensor(5.2553, device='cuda:0', grad_fn=<NllLossBackward0>) 17200 tensor(5.0288, device='cuda:0', grad_fn=<NllLossBackward0>) 17300 tensor(4.8915, device='cuda:0', grad_fn=<NllLossBackward0>) 17400 tensor(4.6779, device='cuda:0', grad_fn=<NllLossBackward0>) 17500 tensor(5.4898, device='cuda:0', grad_fn=<NllLossBackward0>) 17600 tensor(5.1083, device='cuda:0', grad_fn=<NllLossBackward0>) 17700 tensor(4.6789, device='cuda:0', grad_fn=<NllLossBackward0>) 17800 tensor(5.3441, device='cuda:0', grad_fn=<NllLossBackward0>) 17900 tensor(5.1731, device='cuda:0', grad_fn=<NllLossBackward0>) 18000 tensor(5.2611, device='cuda:0', grad_fn=<NllLossBackward0>) 18100 tensor(4.5288, device='cuda:0', grad_fn=<NllLossBackward0>) 18200 tensor(5.5324, device='cuda:0', grad_fn=<NllLossBackward0>) 18300 tensor(4.8522, device='cuda:0', grad_fn=<NllLossBackward0>) 18400 tensor(5.2299, device='cuda:0', grad_fn=<NllLossBackward0>) 18500 tensor(5.1332, device='cuda:0', grad_fn=<NllLossBackward0>) 18600 tensor(5.1331, device='cuda:0', grad_fn=<NllLossBackward0>) 18700 tensor(5.0706, device='cuda:0', grad_fn=<NllLossBackward0>) 18800 tensor(4.9425, device='cuda:0', grad_fn=<NllLossBackward0>) 18900 tensor(5.0168, device='cuda:0', grad_fn=<NllLossBackward0>) 19000 tensor(5.1194, device='cuda:0', grad_fn=<NllLossBackward0>) 19100 tensor(5.1465, device='cuda:0', grad_fn=<NllLossBackward0>) 19200 tensor(5.4118, device='cuda:0', grad_fn=<NllLossBackward0>) 19300 tensor(5.1087, device='cuda:0', grad_fn=<NllLossBackward0>) 19400 tensor(5.2879, device='cuda:0', grad_fn=<NllLossBackward0>) 19500 tensor(4.7952, device='cuda:0', grad_fn=<NllLossBackward0>) 19600 tensor(4.9889, device='cuda:0', grad_fn=<NllLossBackward0>) 19700 tensor(5.2982, device='cuda:0', grad_fn=<NllLossBackward0>) 19800 tensor(4.9682, device='cuda:0', grad_fn=<NllLossBackward0>) 19900 tensor(5.0727, device='cuda:0', grad_fn=<NllLossBackward0>) 20000 tensor(5.3921, device='cuda:0', grad_fn=<NllLossBackward0>) 20100 tensor(5.3464, device='cuda:0', grad_fn=<NllLossBackward0>) 20200 tensor(5.2613, device='cuda:0', grad_fn=<NllLossBackward0>) 20300 tensor(5.2227, device='cuda:0', grad_fn=<NllLossBackward0>) 20400 tensor(4.8486, device='cuda:0', grad_fn=<NllLossBackward0>) 20500 tensor(5.5216, device='cuda:0', grad_fn=<NllLossBackward0>) 20600 tensor(5.1865, device='cuda:0', grad_fn=<NllLossBackward0>) 20700 tensor(5.0252, device='cuda:0', grad_fn=<NllLossBackward0>) 20800 tensor(5.1742, device='cuda:0', grad_fn=<NllLossBackward0>) 20900 tensor(4.9608, device='cuda:0', grad_fn=<NllLossBackward0>) 21000 tensor(4.8754, device='cuda:0', grad_fn=<NllLossBackward0>) 21100 tensor(5.2599, device='cuda:0', grad_fn=<NllLossBackward0>) 21200 tensor(4.7691, device='cuda:0', grad_fn=<NllLossBackward0>) 21300 tensor(5.1363, device='cuda:0', grad_fn=<NllLossBackward0>) 21400 tensor(5.2940, device='cuda:0', grad_fn=<NllLossBackward0>) 21500 tensor(4.8573, device='cuda:0', grad_fn=<NllLossBackward0>) 21600 tensor(5.2410, device='cuda:0', grad_fn=<NllLossBackward0>) 21700 tensor(5.2941, device='cuda:0', grad_fn=<NllLossBackward0>) 21800 tensor(5.0294, device='cuda:0', grad_fn=<NllLossBackward0>) 21900 tensor(5.1803, device='cuda:0', grad_fn=<NllLossBackward0>) 22000 tensor(4.9386, device='cuda:0', grad_fn=<NllLossBackward0>) 22100 tensor(5.3389, device='cuda:0', grad_fn=<NllLossBackward0>) 22200 tensor(5.1924, device='cuda:0', grad_fn=<NllLossBackward0>) 22300 tensor(5.1488, device='cuda:0', grad_fn=<NllLossBackward0>) 22400 tensor(5.4211, device='cuda:0', grad_fn=<NllLossBackward0>) 22500 tensor(4.7259, device='cuda:0', grad_fn=<NllLossBackward0>) 22600 tensor(5.2168, device='cuda:0', grad_fn=<NllLossBackward0>) 22700 tensor(4.9509, device='cuda:0', grad_fn=<NllLossBackward0>) 22800 tensor(5.1694, device='cuda:0', grad_fn=<NllLossBackward0>) 22900 tensor(4.9012, device='cuda:0', grad_fn=<NllLossBackward0>) 23000 tensor(5.2488, device='cuda:0', grad_fn=<NllLossBackward0>) 23100 tensor(5.4762, device='cuda:0', grad_fn=<NllLossBackward0>) 23200 tensor(5.3331, device='cuda:0', grad_fn=<NllLossBackward0>) 23300 tensor(5.1007, device='cuda:0', grad_fn=<NllLossBackward0>) 23400 tensor(5.1397, device='cuda:0', grad_fn=<NllLossBackward0>) 23500 tensor(4.9862, device='cuda:0', grad_fn=<NllLossBackward0>) 23600 tensor(4.5681, device='cuda:0', grad_fn=<NllLossBackward0>) 23700 tensor(5.0539, device='cuda:0', grad_fn=<NllLossBackward0>) 23800 tensor(5.2384, device='cuda:0', grad_fn=<NllLossBackward0>) 23900 tensor(5.2213, device='cuda:0', grad_fn=<NllLossBackward0>) 24000 tensor(5.2034, device='cuda:0', grad_fn=<NllLossBackward0>) 24100 tensor(4.8727, device='cuda:0', grad_fn=<NllLossBackward0>) 24200 tensor(5.1615, device='cuda:0', grad_fn=<NllLossBackward0>) 24300 tensor(5.2397, device='cuda:0', grad_fn=<NllLossBackward0>) 24400 tensor(4.9909, device='cuda:0', grad_fn=<NllLossBackward0>) 24500 tensor(5.3016, device='cuda:0', grad_fn=<NllLossBackward0>) 24600 tensor(4.8689, device='cuda:0', grad_fn=<NllLossBackward0>) 24700 tensor(5.0979, device='cuda:0', grad_fn=<NllLossBackward0>) 24800 tensor(4.8190, device='cuda:0', grad_fn=<NllLossBackward0>) 24900 tensor(5.1405, device='cuda:0', grad_fn=<NllLossBackward0>) 25000 tensor(5.2768, device='cuda:0', grad_fn=<NllLossBackward0>) 25100 tensor(5.1046, device='cuda:0', grad_fn=<NllLossBackward0>) 25200 tensor(5.0897, device='cuda:0', grad_fn=<NllLossBackward0>) 25300 tensor(5.0305, device='cuda:0', grad_fn=<NllLossBackward0>) 25400 tensor(4.8187, device='cuda:0', grad_fn=<NllLossBackward0>) 25500 tensor(4.7033, device='cuda:0', grad_fn=<NllLossBackward0>) 25600 tensor(5.4625, device='cuda:0', grad_fn=<NllLossBackward0>) 25700 tensor(5.0582, device='cuda:0', grad_fn=<NllLossBackward0>) 25800 tensor(4.7515, device='cuda:0', grad_fn=<NllLossBackward0>) 25900 tensor(4.9780, device='cuda:0', grad_fn=<NllLossBackward0>) 26000 tensor(5.0385, device='cuda:0', grad_fn=<NllLossBackward0>) 26100 tensor(5.3317, device='cuda:0', grad_fn=<NllLossBackward0>) 26200 tensor(5.2198, device='cuda:0', grad_fn=<NllLossBackward0>) 26300 tensor(5.2683, device='cuda:0', grad_fn=<NllLossBackward0>) 26400 tensor(4.9987, device='cuda:0', grad_fn=<NllLossBackward0>) 26500 tensor(5.0611, device='cuda:0', grad_fn=<NllLossBackward0>) 26600 tensor(5.2856, device='cuda:0', grad_fn=<NllLossBackward0>) 26700 tensor(5.0887, device='cuda:0', grad_fn=<NllLossBackward0>) 26800 tensor(5.1377, device='cuda:0', grad_fn=<NllLossBackward0>) 26900 tensor(4.5615, device='cuda:0', grad_fn=<NllLossBackward0>) 27000 tensor(5.2209, device='cuda:0', grad_fn=<NllLossBackward0>) 27100 tensor(5.0862, device='cuda:0', grad_fn=<NllLossBackward0>) 27200 tensor(5.0038, device='cuda:0', grad_fn=<NllLossBackward0>) 27300 tensor(5.0321, device='cuda:0', grad_fn=<NllLossBackward0>) 27400 tensor(5.0214, device='cuda:0', grad_fn=<NllLossBackward0>) 27500 tensor(5.2483, device='cuda:0', grad_fn=<NllLossBackward0>) 27600 tensor(4.8013, device='cuda:0', grad_fn=<NllLossBackward0>) 27700 tensor(5.0453, device='cuda:0', grad_fn=<NllLossBackward0>) 27800 tensor(4.8031, device='cuda:0', grad_fn=<NllLossBackward0>) 27900 tensor(4.9852, device='cuda:0', grad_fn=<NllLossBackward0>) 28000 tensor(4.8680, device='cuda:0', grad_fn=<NllLossBackward0>) 28100 tensor(5.3965, device='cuda:0', grad_fn=<NllLossBackward0>) 28200 tensor(5.1008, device='cuda:0', grad_fn=<NllLossBackward0>) 28300 tensor(5.0263, device='cuda:0', grad_fn=<NllLossBackward0>) 28400 tensor(4.9622, device='cuda:0', grad_fn=<NllLossBackward0>) 28500 tensor(5.0803, device='cuda:0', grad_fn=<NllLossBackward0>) 28600 tensor(5.1089, device='cuda:0', grad_fn=<NllLossBackward0>) 28700 tensor(5.6485, device='cuda:0', grad_fn=<NllLossBackward0>) 28800 tensor(5.0195, device='cuda:0', grad_fn=<NllLossBackward0>) 28900 tensor(4.7143, device='cuda:0', grad_fn=<NllLossBackward0>) 29000 tensor(5.2968, device='cuda:0', grad_fn=<NllLossBackward0>) 29100 tensor(5.0881, device='cuda:0', grad_fn=<NllLossBackward0>) 29200 tensor(4.7767, device='cuda:0', grad_fn=<NllLossBackward0>) 29300 tensor(4.9015, device='cuda:0', grad_fn=<NllLossBackward0>) 29400 tensor(5.1177, device='cuda:0', grad_fn=<NllLossBackward0>) 29500 tensor(5.0024, device='cuda:0', grad_fn=<NllLossBackward0>) 29600 tensor(5.0116, device='cuda:0', grad_fn=<NllLossBackward0>) 29700 tensor(5.0913, device='cuda:0', grad_fn=<NllLossBackward0>) 29800 tensor(5.0730, device='cuda:0', grad_fn=<NllLossBackward0>) 29900 tensor(5.2098, device='cuda:0', grad_fn=<NllLossBackward0>) 30000 tensor(4.8949, device='cuda:0', grad_fn=<NllLossBackward0>) 30100 tensor(5.1074, device='cuda:0', grad_fn=<NllLossBackward0>) 30200 tensor(5.2501, device='cuda:0', grad_fn=<NllLossBackward0>) 30300 tensor(5.0952, device='cuda:0', grad_fn=<NllLossBackward0>) 30400 tensor(4.7803, device='cuda:0', grad_fn=<NllLossBackward0>) 30500 tensor(5.1438, device='cuda:0', grad_fn=<NllLossBackward0>) 30600 tensor(5.1865, device='cuda:0', grad_fn=<NllLossBackward0>) 30700 tensor(5.0876, device='cuda:0', grad_fn=<NllLossBackward0>) 30800 tensor(4.9957, device='cuda:0', grad_fn=<NllLossBackward0>) 30900 tensor(4.9111, device='cuda:0', grad_fn=<NllLossBackward0>) 31000 tensor(4.9795, device='cuda:0', grad_fn=<NllLossBackward0>) 31100 tensor(5.1215, device='cuda:0', grad_fn=<NllLossBackward0>) 31200 tensor(4.9652, device='cuda:0', grad_fn=<NllLossBackward0>) 31300 tensor(5.0836, device='cuda:0', grad_fn=<NllLossBackward0>) 31400 tensor(4.7480, device='cuda:0', grad_fn=<NllLossBackward0>) 31500 tensor(5.0733, device='cuda:0', grad_fn=<NllLossBackward0>) 31600 tensor(5.0257, device='cuda:0', grad_fn=<NllLossBackward0>) 31700 tensor(5.2194, device='cuda:0', grad_fn=<NllLossBackward0>) 31800 tensor(4.6968, device='cuda:0', grad_fn=<NllLossBackward0>) 31900 tensor(4.8459, device='cuda:0', grad_fn=<NllLossBackward0>) 32000 tensor(4.9492, device='cuda:0', grad_fn=<NllLossBackward0>) 32100 tensor(5.1559, device='cuda:0', grad_fn=<NllLossBackward0>) 32200 tensor(4.9369, device='cuda:0', grad_fn=<NllLossBackward0>) 32300 tensor(5.1198, device='cuda:0', grad_fn=<NllLossBackward0>) 32400 tensor(4.9705, device='cuda:0', grad_fn=<NllLossBackward0>) 32500 tensor(5.1361, device='cuda:0', grad_fn=<NllLossBackward0>) 32600 tensor(4.9825, device='cuda:0', grad_fn=<NllLossBackward0>) 32700 tensor(5.4973, device='cuda:0', grad_fn=<NllLossBackward0>) 32800 tensor(5.1503, device='cuda:0', grad_fn=<NllLossBackward0>) 32900 tensor(5.0624, device='cuda:0', grad_fn=<NllLossBackward0>) 33000 tensor(5.2216, device='cuda:0', grad_fn=<NllLossBackward0>) 33100 tensor(4.9986, device='cuda:0', grad_fn=<NllLossBackward0>) 33200 tensor(5.1666, device='cuda:0', grad_fn=<NllLossBackward0>) 33300 tensor(5.0774, device='cuda:0', grad_fn=<NllLossBackward0>) 33400 tensor(4.7154, device='cuda:0', grad_fn=<NllLossBackward0>) 33500 tensor(4.8050, device='cuda:0', grad_fn=<NllLossBackward0>) 33600 tensor(5.0121, device='cuda:0', grad_fn=<NllLossBackward0>) 33700 tensor(4.9336, device='cuda:0', grad_fn=<NllLossBackward0>) 33800 tensor(4.9983, device='cuda:0', grad_fn=<NllLossBackward0>) 33900 tensor(4.9299, device='cuda:0', grad_fn=<NllLossBackward0>) 34000 tensor(4.8085, device='cuda:0', grad_fn=<NllLossBackward0>) 34100 tensor(4.6787, device='cuda:0', grad_fn=<NllLossBackward0>) 34200 tensor(5.0441, device='cuda:0', grad_fn=<NllLossBackward0>) 34300 tensor(5.0500, device='cuda:0', grad_fn=<NllLossBackward0>) 34400 tensor(4.8602, device='cuda:0', grad_fn=<NllLossBackward0>) 34500 tensor(5.1163, device='cuda:0', grad_fn=<NllLossBackward0>) 34600 tensor(5.0843, device='cuda:0', grad_fn=<NllLossBackward0>) 34700 tensor(5.1913, device='cuda:0', grad_fn=<NllLossBackward0>) 34800 tensor(5.0463, device='cuda:0', grad_fn=<NllLossBackward0>) 34900 tensor(5.0379, device='cuda:0', grad_fn=<NllLossBackward0>) 35000 tensor(5.1605, device='cuda:0', grad_fn=<NllLossBackward0>) 35100 tensor(4.9475, device='cuda:0', grad_fn=<NllLossBackward0>) 35200 tensor(4.7895, device='cuda:0', grad_fn=<NllLossBackward0>) 35300 tensor(4.9079, device='cuda:0', grad_fn=<NllLossBackward0>) 35400 tensor(4.8474, device='cuda:0', grad_fn=<NllLossBackward0>) 35500 tensor(5.1182, device='cuda:0', grad_fn=<NllLossBackward0>) 35600 tensor(4.9876, device='cuda:0', grad_fn=<NllLossBackward0>) 35700 tensor(5.0740, device='cuda:0', grad_fn=<NllLossBackward0>) 35800 tensor(5.0899, device='cuda:0', grad_fn=<NllLossBackward0>) 35900 tensor(5.4795, device='cuda:0', grad_fn=<NllLossBackward0>) 36000 tensor(4.8221, device='cuda:0', grad_fn=<NllLossBackward0>) 36100 tensor(5.2199, device='cuda:0', grad_fn=<NllLossBackward0>) 36200 tensor(5.1585, device='cuda:0', grad_fn=<NllLossBackward0>) 36300 tensor(4.9132, device='cuda:0', grad_fn=<NllLossBackward0>) 36400 tensor(4.6865, device='cuda:0', grad_fn=<NllLossBackward0>) 36500 tensor(5.0982, device='cuda:0', grad_fn=<NllLossBackward0>) 36600 tensor(5.1886, device='cuda:0', grad_fn=<NllLossBackward0>) 36700 tensor(5.2310, device='cuda:0', grad_fn=<NllLossBackward0>) 36800 tensor(4.8729, device='cuda:0', grad_fn=<NllLossBackward0>) 36900 tensor(5.1362, device='cuda:0', grad_fn=<NllLossBackward0>) 37000 tensor(5.2179, device='cuda:0', grad_fn=<NllLossBackward0>) 37100 tensor(5.0787, device='cuda:0', grad_fn=<NllLossBackward0>) 37200 tensor(5.1929, device='cuda:0', grad_fn=<NllLossBackward0>) 37300 tensor(5.1267, device='cuda:0', grad_fn=<NllLossBackward0>) 37400 tensor(4.8166, device='cuda:0', grad_fn=<NllLossBackward0>) 37500 tensor(4.7019, device='cuda:0', grad_fn=<NllLossBackward0>) 37600 tensor(5.0429, device='cuda:0', grad_fn=<NllLossBackward0>) 37700 tensor(4.9993, device='cuda:0', grad_fn=<NllLossBackward0>) 37800 tensor(4.9434, device='cuda:0', grad_fn=<NllLossBackward0>) 37900 tensor(4.9565, device='cuda:0', grad_fn=<NllLossBackward0>) 38000 tensor(5.0422, device='cuda:0', grad_fn=<NllLossBackward0>) 38100 tensor(4.9607, device='cuda:0', grad_fn=<NllLossBackward0>) 38200 tensor(5.2120, device='cuda:0', grad_fn=<NllLossBackward0>) 38300 tensor(5.1508, device='cuda:0', grad_fn=<NllLossBackward0>) 38400 tensor(5.2408, device='cuda:0', grad_fn=<NllLossBackward0>) 38500 tensor(5.1387, device='cuda:0', grad_fn=<NllLossBackward0>) 38600 tensor(4.9913, device='cuda:0', grad_fn=<NllLossBackward0>) 38700 tensor(5.0389, device='cuda:0', grad_fn=<NllLossBackward0>) 38800 tensor(5.2357, device='cuda:0', grad_fn=<NllLossBackward0>) 38900 tensor(5.4844, device='cuda:0', grad_fn=<NllLossBackward0>) 39000 tensor(4.8261, device='cuda:0', grad_fn=<NllLossBackward0>) 39100 tensor(5.0514, device='cuda:0', grad_fn=<NllLossBackward0>) 39200 tensor(5.1650, device='cuda:0', grad_fn=<NllLossBackward0>) 39300 tensor(4.9774, device='cuda:0', grad_fn=<NllLossBackward0>) 39400 tensor(5.0853, device='cuda:0', grad_fn=<NllLossBackward0>) 39500 tensor(5.1007, device='cuda:0', grad_fn=<NllLossBackward0>) 39600 tensor(5.2591, device='cuda:0', grad_fn=<NllLossBackward0>) 39700 tensor(5.0162, device='cuda:0', grad_fn=<NllLossBackward0>) 39800 tensor(5.1029, device='cuda:0', grad_fn=<NllLossBackward0>) 39900 tensor(4.8540, device='cuda:0', grad_fn=<NllLossBackward0>) 40000 tensor(5.0950, device='cuda:0', grad_fn=<NllLossBackward0>) 40100 tensor(5.1870, device='cuda:0', grad_fn=<NllLossBackward0>) 40200 tensor(4.8280, device='cuda:0', grad_fn=<NllLossBackward0>) 40300 tensor(5.0794, device='cuda:0', grad_fn=<NllLossBackward0>) 40400 tensor(5.0300, device='cuda:0', grad_fn=<NllLossBackward0>) 40500 tensor(5.3105, device='cuda:0', grad_fn=<NllLossBackward0>) 40600 tensor(5.1316, device='cuda:0', grad_fn=<NllLossBackward0>) 40700 tensor(5.0367, device='cuda:0', grad_fn=<NllLossBackward0>) 40800 tensor(4.9724, device='cuda:0', grad_fn=<NllLossBackward0>) 40900 tensor(5.0658, device='cuda:0', grad_fn=<NllLossBackward0>) 41000 tensor(4.9169, device='cuda:0', grad_fn=<NllLossBackward0>) 41100 tensor(4.4952, device='cuda:0', grad_fn=<NllLossBackward0>) 41200 tensor(4.8536, device='cuda:0', grad_fn=<NllLossBackward0>) 41300 tensor(5.1255, device='cuda:0', grad_fn=<NllLossBackward0>) 41400 tensor(5.1006, device='cuda:0', grad_fn=<NllLossBackward0>) 41500 tensor(4.9304, device='cuda:0', grad_fn=<NllLossBackward0>) 41600 tensor(4.8720, device='cuda:0', grad_fn=<NllLossBackward0>) 41700 tensor(5.1611, device='cuda:0', grad_fn=<NllLossBackward0>) 41800 tensor(4.9487, device='cuda:0', grad_fn=<NllLossBackward0>) 41900 tensor(4.7736, device='cuda:0', grad_fn=<NllLossBackward0>) 42000 tensor(5.2250, device='cuda:0', grad_fn=<NllLossBackward0>) 42100 tensor(5.3416, device='cuda:0', grad_fn=<NllLossBackward0>) 42200 tensor(4.9317, device='cuda:0', grad_fn=<NllLossBackward0>) 42300 tensor(4.8594, device='cuda:0', grad_fn=<NllLossBackward0>) 42400 tensor(4.7213, device='cuda:0', grad_fn=<NllLossBackward0>) 42500 tensor(4.9424, device='cuda:0', grad_fn=<NllLossBackward0>) 42600 tensor(4.9085, device='cuda:0', grad_fn=<NllLossBackward0>) 42700 tensor(5.0864, device='cuda:0', grad_fn=<NllLossBackward0>) 42800 tensor(4.5522, device='cuda:0', grad_fn=<NllLossBackward0>) 42900 tensor(5.4337, device='cuda:0', grad_fn=<NllLossBackward0>) 43000 tensor(5.0700, device='cuda:0', grad_fn=<NllLossBackward0>) 43100 tensor(5.1494, device='cuda:0', grad_fn=<NllLossBackward0>) 43200 tensor(4.7431, device='cuda:0', grad_fn=<NllLossBackward0>) 43300 tensor(5.3639, device='cuda:0', grad_fn=<NllLossBackward0>) 43400 tensor(4.8963, device='cuda:0', grad_fn=<NllLossBackward0>) 43500 tensor(5.1136, device='cuda:0', grad_fn=<NllLossBackward0>) 43600 tensor(4.8048, device='cuda:0', grad_fn=<NllLossBackward0>) 43700 tensor(4.9236, device='cuda:0', grad_fn=<NllLossBackward0>) 43800 tensor(4.9797, device='cuda:0', grad_fn=<NllLossBackward0>) 43900 tensor(5.0236, device='cuda:0', grad_fn=<NllLossBackward0>) 44000 tensor(5.0134, device='cuda:0', grad_fn=<NllLossBackward0>) 44100 tensor(4.8278, device='cuda:0', grad_fn=<NllLossBackward0>) 44200 tensor(4.7282, device='cuda:0', grad_fn=<NllLossBackward0>) 44300 tensor(5.1968, device='cuda:0', grad_fn=<NllLossBackward0>) 44400 tensor(4.9318, device='cuda:0', grad_fn=<NllLossBackward0>) 44500 tensor(4.9458, device='cuda:0', grad_fn=<NllLossBackward0>) 44600 tensor(5.0479, device='cuda:0', grad_fn=<NllLossBackward0>) 44700 tensor(4.8669, device='cuda:0', grad_fn=<NllLossBackward0>) 44800 tensor(5.1006, device='cuda:0', grad_fn=<NllLossBackward0>) 44900 tensor(5.0761, device='cuda:0', grad_fn=<NllLossBackward0>) 45000 tensor(4.8570, device='cuda:0', grad_fn=<NllLossBackward0>) 45100 tensor(4.9536, device='cuda:0', grad_fn=<NllLossBackward0>) 45200 tensor(4.7665, device='cuda:0', grad_fn=<NllLossBackward0>) 45300 tensor(5.1462, device='cuda:0', grad_fn=<NllLossBackward0>) 45400 tensor(4.7666, device='cuda:0', grad_fn=<NllLossBackward0>) 45500 tensor(4.9350, device='cuda:0', grad_fn=<NllLossBackward0>) 45600 tensor(4.9407, device='cuda:0', grad_fn=<NllLossBackward0>) 45700 tensor(4.9847, device='cuda:0', grad_fn=<NllLossBackward0>) 45800 tensor(4.3695, device='cuda:0', grad_fn=<NllLossBackward0>) 45900 tensor(5.0090, device='cuda:0', grad_fn=<NllLossBackward0>) 46000 tensor(4.5963, device='cuda:0', grad_fn=<NllLossBackward0>) 46100 tensor(5.2931, device='cuda:0', grad_fn=<NllLossBackward0>) 46200 tensor(5.0931, device='cuda:0', grad_fn=<NllLossBackward0>) 46300 tensor(4.8944, device='cuda:0', grad_fn=<NllLossBackward0>) 46400 tensor(5.1315, device='cuda:0', grad_fn=<NllLossBackward0>) 46500 tensor(5.0825, device='cuda:0', grad_fn=<NllLossBackward0>) 46600 tensor(4.8329, device='cuda:0', grad_fn=<NllLossBackward0>) 46700 tensor(5.1103, device='cuda:0', grad_fn=<NllLossBackward0>) 46800 tensor(5.0603, device='cuda:0', grad_fn=<NllLossBackward0>) 46900 tensor(5.1512, device='cuda:0', grad_fn=<NllLossBackward0>) 47000 tensor(4.6868, device='cuda:0', grad_fn=<NllLossBackward0>) 47100 tensor(4.8712, device='cuda:0', grad_fn=<NllLossBackward0>) 47200 tensor(5.1912, device='cuda:0', grad_fn=<NllLossBackward0>) 47300 tensor(5.0159, device='cuda:0', grad_fn=<NllLossBackward0>) 47400 tensor(5.0270, device='cuda:0', grad_fn=<NllLossBackward0>) 47500 tensor(4.6956, device='cuda:0', grad_fn=<NllLossBackward0>) 47600 tensor(4.9921, device='cuda:0', grad_fn=<NllLossBackward0>) 47700 tensor(5.1164, device='cuda:0', grad_fn=<NllLossBackward0>) 47800 tensor(4.7971, device='cuda:0', grad_fn=<NllLossBackward0>) 47900 tensor(4.9729, device='cuda:0', grad_fn=<NllLossBackward0>) 48000 tensor(5.1074, device='cuda:0', grad_fn=<NllLossBackward0>) 48100 tensor(5.0341, device='cuda:0', grad_fn=<NllLossBackward0>) 48200 tensor(4.8993, device='cuda:0', grad_fn=<NllLossBackward0>) 48300 tensor(4.8756, device='cuda:0', grad_fn=<NllLossBackward0>) 48400 tensor(5.1059, device='cuda:0', grad_fn=<NllLossBackward0>) 48500 tensor(5.0636, device='cuda:0', grad_fn=<NllLossBackward0>) 48600 tensor(4.9931, device='cuda:0', grad_fn=<NllLossBackward0>) 48700 tensor(4.9768, device='cuda:0', grad_fn=<NllLossBackward0>) 48800 tensor(5.1949, device='cuda:0', grad_fn=<NllLossBackward0>) 48900 tensor(4.7792, device='cuda:0', grad_fn=<NllLossBackward0>) 49000 tensor(4.7743, device='cuda:0', grad_fn=<NllLossBackward0>) 49100 tensor(5.3373, device='cuda:0', grad_fn=<NllLossBackward0>) 49200 tensor(4.7696, device='cuda:0', grad_fn=<NllLossBackward0>) 49300 tensor(4.7552, device='cuda:0', grad_fn=<NllLossBackward0>) 49400 tensor(4.8440, device='cuda:0', grad_fn=<NllLossBackward0>) 49500 tensor(4.9257, device='cuda:0', grad_fn=<NllLossBackward0>) 49600 tensor(5.1038, device='cuda:0', grad_fn=<NllLossBackward0>) 49700 tensor(4.9334, device='cuda:0', grad_fn=<NllLossBackward0>) 49800 tensor(5.1488, device='cuda:0', grad_fn=<NllLossBackward0>) 49900 tensor(5.1520, device='cuda:0', grad_fn=<NllLossBackward0>) 50000 tensor(4.6342, device='cuda:0', grad_fn=<NllLossBackward0>) 50100 tensor(5.1441, device='cuda:0', grad_fn=<NllLossBackward0>) 50200 tensor(4.4512, device='cuda:0', grad_fn=<NllLossBackward0>) 50300 tensor(5.0929, device='cuda:0', grad_fn=<NllLossBackward0>) 50400 tensor(5.1754, device='cuda:0', grad_fn=<NllLossBackward0>) 50500 tensor(4.9708, device='cuda:0', grad_fn=<NllLossBackward0>) 50600 tensor(4.7339, device='cuda:0', grad_fn=<NllLossBackward0>) 50700 tensor(4.8156, device='cuda:0', grad_fn=<NllLossBackward0>) 50800 tensor(5.0074, device='cuda:0', grad_fn=<NllLossBackward0>) 50900 tensor(4.7649, device='cuda:0', grad_fn=<NllLossBackward0>) 51000 tensor(5.0744, device='cuda:0', grad_fn=<NllLossBackward0>) 51100 tensor(5.0727, device='cuda:0', grad_fn=<NllLossBackward0>) 51200 tensor(4.8884, device='cuda:0', grad_fn=<NllLossBackward0>) 51300 tensor(5.0004, device='cuda:0', grad_fn=<NllLossBackward0>) 51400 tensor(4.7368, device='cuda:0', grad_fn=<NllLossBackward0>) 51500 tensor(4.9471, device='cuda:0', grad_fn=<NllLossBackward0>) 51600 tensor(5.0216, device='cuda:0', grad_fn=<NllLossBackward0>) 51700 tensor(4.8408, device='cuda:0', grad_fn=<NllLossBackward0>) 51800 tensor(4.7706, device='cuda:0', grad_fn=<NllLossBackward0>) 51900 tensor(5.1493, device='cuda:0', grad_fn=<NllLossBackward0>) 52000 tensor(4.9272, device='cuda:0', grad_fn=<NllLossBackward0>) 52100 tensor(4.8667, device='cuda:0', grad_fn=<NllLossBackward0>) 52200 tensor(4.8331, device='cuda:0', grad_fn=<NllLossBackward0>) 52300 tensor(4.9672, device='cuda:0', grad_fn=<NllLossBackward0>) 52400 tensor(4.9821, device='cuda:0', grad_fn=<NllLossBackward0>) 52500 tensor(5.1103, device='cuda:0', grad_fn=<NllLossBackward0>) 52600 tensor(5.0463, device='cuda:0', grad_fn=<NllLossBackward0>) 52700 tensor(5.0696, device='cuda:0', grad_fn=<NllLossBackward0>) 52800 tensor(4.5670, device='cuda:0', grad_fn=<NllLossBackward0>) 52900 tensor(4.9175, device='cuda:0', grad_fn=<NllLossBackward0>) 53000 tensor(5.0619, device='cuda:0', grad_fn=<NllLossBackward0>) 53100 tensor(4.9976, device='cuda:0', grad_fn=<NllLossBackward0>) 53200 tensor(4.6688, device='cuda:0', grad_fn=<NllLossBackward0>) 53300 tensor(4.9201, device='cuda:0', grad_fn=<NllLossBackward0>) 53400 tensor(4.7648, device='cuda:0', grad_fn=<NllLossBackward0>) 53500 tensor(5.0633, device='cuda:0', grad_fn=<NllLossBackward0>) 53600 tensor(5.0900, device='cuda:0', grad_fn=<NllLossBackward0>) 53700 tensor(4.6773, device='cuda:0', grad_fn=<NllLossBackward0>) 53800 tensor(4.9783, device='cuda:0', grad_fn=<NllLossBackward0>) 53900 tensor(4.8892, device='cuda:0', grad_fn=<NllLossBackward0>) 54000 tensor(5.1447, device='cuda:0', grad_fn=<NllLossBackward0>) 54100 tensor(4.7897, device='cuda:0', grad_fn=<NllLossBackward0>) 54200 tensor(4.9532, device='cuda:0', grad_fn=<NllLossBackward0>) 54300 tensor(5.0852, device='cuda:0', grad_fn=<NllLossBackward0>) 54400 tensor(4.8126, device='cuda:0', grad_fn=<NllLossBackward0>) 54500 tensor(4.7961, device='cuda:0', grad_fn=<NllLossBackward0>) 54600 tensor(4.4173, device='cuda:0', grad_fn=<NllLossBackward0>) 54700 tensor(5.0556, device='cuda:0', grad_fn=<NllLossBackward0>) 54800 tensor(4.8861, device='cuda:0', grad_fn=<NllLossBackward0>) 54900 tensor(4.9520, device='cuda:0', grad_fn=<NllLossBackward0>) 55000 tensor(4.9933, device='cuda:0', grad_fn=<NllLossBackward0>) 55100 tensor(5.1774, device='cuda:0', grad_fn=<NllLossBackward0>) 55200 tensor(4.8481, device='cuda:0', grad_fn=<NllLossBackward0>) 55300 tensor(4.9836, device='cuda:0', grad_fn=<NllLossBackward0>) 55400 tensor(4.7960, device='cuda:0', grad_fn=<NllLossBackward0>) 55500 tensor(5.0643, device='cuda:0', grad_fn=<NllLossBackward0>) 55600 tensor(5.0122, device='cuda:0', grad_fn=<NllLossBackward0>) 55700 tensor(4.7390, device='cuda:0', grad_fn=<NllLossBackward0>) 55800 tensor(5.1234, device='cuda:0', grad_fn=<NllLossBackward0>) 55900 tensor(4.4134, device='cuda:0', grad_fn=<NllLossBackward0>) 56000 tensor(4.0267, device='cuda:0', grad_fn=<NllLossBackward0>) 56100 tensor(4.9464, device='cuda:0', grad_fn=<NllLossBackward0>) 56200 tensor(4.9766, device='cuda:0', grad_fn=<NllLossBackward0>) 56300 tensor(4.6557, device='cuda:0', grad_fn=<NllLossBackward0>) 56400 tensor(4.9807, device='cuda:0', grad_fn=<NllLossBackward0>) 56500 tensor(5.0741, device='cuda:0', grad_fn=<NllLossBackward0>) 56600 tensor(5.2151, device='cuda:0', grad_fn=<NllLossBackward0>) 56700 tensor(4.7876, device='cuda:0', grad_fn=<NllLossBackward0>) 56800 tensor(4.8416, device='cuda:0', grad_fn=<NllLossBackward0>) 56900 tensor(4.6740, device='cuda:0', grad_fn=<NllLossBackward0>) 57000 tensor(4.9346, device='cuda:0', grad_fn=<NllLossBackward0>) 57100 tensor(4.4737, device='cuda:0', grad_fn=<NllLossBackward0>) 57200 tensor(4.6889, device='cuda:0', grad_fn=<NllLossBackward0>) 57300 tensor(4.6792, device='cuda:0', grad_fn=<NllLossBackward0>) 57400 tensor(4.8290, device='cuda:0', grad_fn=<NllLossBackward0>) 57500 tensor(4.8428, device='cuda:0', grad_fn=<NllLossBackward0>) 57600 tensor(4.7810, device='cuda:0', grad_fn=<NllLossBackward0>) 57700 tensor(4.9318, device='cuda:0', grad_fn=<NllLossBackward0>) 57800 tensor(5.0835, device='cuda:0', grad_fn=<NllLossBackward0>) 57900 tensor(4.4041, device='cuda:0', grad_fn=<NllLossBackward0>) 58000 tensor(4.9021, device='cuda:0', grad_fn=<NllLossBackward0>) 58100 tensor(4.9826, device='cuda:0', grad_fn=<NllLossBackward0>) 58200 tensor(4.8855, device='cuda:0', grad_fn=<NllLossBackward0>) 58300 tensor(5.1050, device='cuda:0', grad_fn=<NllLossBackward0>) 58400 tensor(4.8503, device='cuda:0', grad_fn=<NllLossBackward0>) 58500 tensor(4.9113, device='cuda:0', grad_fn=<NllLossBackward0>) 58600 tensor(4.8079, device='cuda:0', grad_fn=<NllLossBackward0>) 58700 tensor(4.8311, device='cuda:0', grad_fn=<NllLossBackward0>) 58800 tensor(4.8372, device='cuda:0', grad_fn=<NllLossBackward0>) 58900 tensor(5.0136, device='cuda:0', grad_fn=<NllLossBackward0>) 59000 tensor(4.9982, device='cuda:0', grad_fn=<NllLossBackward0>) 59100 tensor(4.9190, device='cuda:0', grad_fn=<NllLossBackward0>) 59200 tensor(4.3102, device='cuda:0', grad_fn=<NllLossBackward0>) 59300 tensor(4.7055, device='cuda:0', grad_fn=<NllLossBackward0>) 59400 tensor(5.0937, device='cuda:0', grad_fn=<NllLossBackward0>) 59500 tensor(4.7516, device='cuda:0', grad_fn=<NllLossBackward0>) 59600 tensor(4.8521, device='cuda:0', grad_fn=<NllLossBackward0>) 59700 tensor(4.8566, device='cuda:0', grad_fn=<NllLossBackward0>) 59800 tensor(4.9476, device='cuda:0', grad_fn=<NllLossBackward0>) 59900 tensor(4.9672, device='cuda:0', grad_fn=<NllLossBackward0>) 60000 tensor(5.1426, device='cuda:0', grad_fn=<NllLossBackward0>) 60100 tensor(4.6298, device='cuda:0', grad_fn=<NllLossBackward0>) 60200 tensor(4.9855, device='cuda:0', grad_fn=<NllLossBackward0>) 60300 tensor(5.1328, device='cuda:0', grad_fn=<NllLossBackward0>) 60400 tensor(4.8736, device='cuda:0', grad_fn=<NllLossBackward0>) 60500 tensor(4.8211, device='cuda:0', grad_fn=<NllLossBackward0>) 60600 tensor(4.6593, device='cuda:0', grad_fn=<NllLossBackward0>) 60700 tensor(4.9276, device='cuda:0', grad_fn=<NllLossBackward0>) 60800 tensor(4.9521, device='cuda:0', grad_fn=<NllLossBackward0>) 60900 tensor(5.1064, device='cuda:0', grad_fn=<NllLossBackward0>) 61000 tensor(4.8707, device='cuda:0', grad_fn=<NllLossBackward0>) 61100 tensor(5.0560, device='cuda:0', grad_fn=<NllLossBackward0>) 61200 tensor(4.9808, device='cuda:0', grad_fn=<NllLossBackward0>) 61300 tensor(4.7554, device='cuda:0', grad_fn=<NllLossBackward0>) 61400 tensor(4.9695, device='cuda:0', grad_fn=<NllLossBackward0>) 61500 tensor(4.9166, device='cuda:0', grad_fn=<NllLossBackward0>) 61600 tensor(5.0340, device='cuda:0', grad_fn=<NllLossBackward0>) 61700 tensor(4.9099, device='cuda:0', grad_fn=<NllLossBackward0>) 61800 tensor(4.7813, device='cuda:0', grad_fn=<NllLossBackward0>) 61900 tensor(5.0149, device='cuda:0', grad_fn=<NllLossBackward0>) 62000 tensor(4.9312, device='cuda:0', grad_fn=<NllLossBackward0>) 62100 tensor(4.7428, device='cuda:0', grad_fn=<NllLossBackward0>) 62200 tensor(4.6879, device='cuda:0', grad_fn=<NllLossBackward0>) 62300 tensor(4.5486, device='cuda:0', grad_fn=<NllLossBackward0>) 62400 tensor(4.8862, device='cuda:0', grad_fn=<NllLossBackward0>) 62500 tensor(4.9385, device='cuda:0', grad_fn=<NllLossBackward0>) 62600 tensor(5.1668, device='cuda:0', grad_fn=<NllLossBackward0>) 62700 tensor(4.7978, device='cuda:0', grad_fn=<NllLossBackward0>) 62800 tensor(5.0571, device='cuda:0', grad_fn=<NllLossBackward0>) 62900 tensor(4.7626, device='cuda:0', grad_fn=<NllLossBackward0>) 63000 tensor(4.6488, device='cuda:0', grad_fn=<NllLossBackward0>) 63100 tensor(4.6918, device='cuda:0', grad_fn=<NllLossBackward0>) 63200 tensor(4.3733, device='cuda:0', grad_fn=<NllLossBackward0>) 63300 tensor(5.0669, device='cuda:0', grad_fn=<NllLossBackward0>) 63400 tensor(4.7177, device='cuda:0', grad_fn=<NllLossBackward0>) 63500 tensor(4.8673, device='cuda:0', grad_fn=<NllLossBackward0>) 63600 tensor(4.9796, device='cuda:0', grad_fn=<NllLossBackward0>) 63700 tensor(5.0491, device='cuda:0', grad_fn=<NllLossBackward0>) 63800 tensor(4.7432, device='cuda:0', grad_fn=<NllLossBackward0>) 63900 tensor(5.0631, device='cuda:0', grad_fn=<NllLossBackward0>) 64000 tensor(4.9573, device='cuda:0', grad_fn=<NllLossBackward0>) 64100 tensor(4.5370, device='cuda:0', grad_fn=<NllLossBackward0>) 64200 tensor(4.9560, device='cuda:0', grad_fn=<NllLossBackward0>) 64300 tensor(4.9266, device='cuda:0', grad_fn=<NllLossBackward0>) 64400 tensor(5.0917, device='cuda:0', grad_fn=<NllLossBackward0>) 64500 tensor(4.9799, device='cuda:0', grad_fn=<NllLossBackward0>) 64600 tensor(5.0494, device='cuda:0', grad_fn=<NllLossBackward0>) 64700 tensor(4.9022, device='cuda:0', grad_fn=<NllLossBackward0>) 64800 tensor(4.8322, device='cuda:0', grad_fn=<NllLossBackward0>) 64900 tensor(4.5922, device='cuda:0', grad_fn=<NllLossBackward0>) 65000 tensor(4.9516, device='cuda:0', grad_fn=<NllLossBackward0>) 65100 tensor(4.8843, device='cuda:0', grad_fn=<NllLossBackward0>) 65200 tensor(5.0609, device='cuda:0', grad_fn=<NllLossBackward0>) 65300 tensor(4.8728, device='cuda:0', grad_fn=<NllLossBackward0>) 65400 tensor(4.9886, device='cuda:0', grad_fn=<NllLossBackward0>) 65500 tensor(4.9286, device='cuda:0', grad_fn=<NllLossBackward0>) 65600 tensor(4.8055, device='cuda:0', grad_fn=<NllLossBackward0>) 65700 tensor(4.9524, device='cuda:0', grad_fn=<NllLossBackward0>) 65800 tensor(4.9742, device='cuda:0', grad_fn=<NllLossBackward0>) 65900 tensor(4.8229, device='cuda:0', grad_fn=<NllLossBackward0>) 66000 tensor(4.8436, device='cuda:0', grad_fn=<NllLossBackward0>) 66100 tensor(4.8835, device='cuda:0', grad_fn=<NllLossBackward0>) 66200 tensor(4.8788, device='cuda:0', grad_fn=<NllLossBackward0>) 66300 tensor(4.3317, device='cuda:0', grad_fn=<NllLossBackward0>) 66400 tensor(4.8368, device='cuda:0', grad_fn=<NllLossBackward0>) 66500 tensor(5.0797, device='cuda:0', grad_fn=<NllLossBackward0>) 66600 tensor(4.8528, device='cuda:0', grad_fn=<NllLossBackward0>) 66700 tensor(5.0445, device='cuda:0', grad_fn=<NllLossBackward0>) 66800 tensor(5.0235, device='cuda:0', grad_fn=<NllLossBackward0>) 66900 tensor(4.8326, device='cuda:0', grad_fn=<NllLossBackward0>) 67000 tensor(4.9944, device='cuda:0', grad_fn=<NllLossBackward0>) 67100 tensor(4.5768, device='cuda:0', grad_fn=<NllLossBackward0>) 67200 tensor(4.7574, device='cuda:0', grad_fn=<NllLossBackward0>) 67300 tensor(4.4299, device='cuda:0', grad_fn=<NllLossBackward0>) 67400 tensor(5.0257, device='cuda:0', grad_fn=<NllLossBackward0>) 67500 tensor(4.9804, device='cuda:0', grad_fn=<NllLossBackward0>) 67600 tensor(4.7536, device='cuda:0', grad_fn=<NllLossBackward0>) 67700 tensor(4.8228, device='cuda:0', grad_fn=<NllLossBackward0>) 67800 tensor(4.9779, device='cuda:0', grad_fn=<NllLossBackward0>) 67900 tensor(5.0636, device='cuda:0', grad_fn=<NllLossBackward0>) 68000 tensor(4.6665, device='cuda:0', grad_fn=<NllLossBackward0>) 68100 tensor(4.7659, device='cuda:0', grad_fn=<NllLossBackward0>) 68200 tensor(4.8760, device='cuda:0', grad_fn=<NllLossBackward0>) 68300 tensor(4.7543, device='cuda:0', grad_fn=<NllLossBackward0>) 68400 tensor(4.9039, device='cuda:0', grad_fn=<NllLossBackward0>) 68500 tensor(4.9760, device='cuda:0', grad_fn=<NllLossBackward0>) 68600 tensor(5.2005, device='cuda:0', grad_fn=<NllLossBackward0>) 68700 tensor(4.9386, device='cuda:0', grad_fn=<NllLossBackward0>) 68800 tensor(5.0149, device='cuda:0', grad_fn=<NllLossBackward0>) 68900 tensor(4.9252, device='cuda:0', grad_fn=<NllLossBackward0>) 69000 tensor(4.9424, device='cuda:0', grad_fn=<NllLossBackward0>) 69100 tensor(4.7407, device='cuda:0', grad_fn=<NllLossBackward0>) 69200 tensor(4.6976, device='cuda:0', grad_fn=<NllLossBackward0>) 69300 tensor(5.0037, device='cuda:0', grad_fn=<NllLossBackward0>) 69400 tensor(4.9396, device='cuda:0', grad_fn=<NllLossBackward0>) 69500 tensor(4.8608, device='cuda:0', grad_fn=<NllLossBackward0>) 69600 tensor(4.8549, device='cuda:0', grad_fn=<NllLossBackward0>) 69700 tensor(4.3331, device='cuda:0', grad_fn=<NllLossBackward0>) 69800 tensor(4.7600, device='cuda:0', grad_fn=<NllLossBackward0>) 69900 tensor(5.1123, device='cuda:0', grad_fn=<NllLossBackward0>) 70000 tensor(4.9554, device='cuda:0', grad_fn=<NllLossBackward0>) 70100 tensor(4.8317, device='cuda:0', grad_fn=<NllLossBackward0>) 70200 tensor(4.6206, device='cuda:0', grad_fn=<NllLossBackward0>) 70300 tensor(5.0007, device='cuda:0', grad_fn=<NllLossBackward0>) 70400 tensor(4.7921, device='cuda:0', grad_fn=<NllLossBackward0>) 70500 tensor(4.9168, device='cuda:0', grad_fn=<NllLossBackward0>) 70600 tensor(4.9162, device='cuda:0', grad_fn=<NllLossBackward0>) 70700 tensor(4.9351, device='cuda:0', grad_fn=<NllLossBackward0>) 70800 tensor(4.7952, device='cuda:0', grad_fn=<NllLossBackward0>) 70900 tensor(4.8172, device='cuda:0', grad_fn=<NllLossBackward0>) 71000 tensor(4.7470, device='cuda:0', grad_fn=<NllLossBackward0>) 71100 tensor(4.8581, device='cuda:0', grad_fn=<NllLossBackward0>) 71200 tensor(4.7746, device='cuda:0', grad_fn=<NllLossBackward0>) 71300 tensor(4.8719, device='cuda:0', grad_fn=<NllLossBackward0>) 71400 tensor(4.9696, device='cuda:0', grad_fn=<NllLossBackward0>) 71500 tensor(4.8571, device='cuda:0', grad_fn=<NllLossBackward0>) 71600 tensor(5.0854, device='cuda:0', grad_fn=<NllLossBackward0>) 71700 tensor(5.0178, device='cuda:0', grad_fn=<NllLossBackward0>) 71800 tensor(5.0655, device='cuda:0', grad_fn=<NllLossBackward0>) 71900 tensor(4.7754, device='cuda:0', grad_fn=<NllLossBackward0>) 72000 tensor(4.8953, device='cuda:0', grad_fn=<NllLossBackward0>) 72100 tensor(4.6920, device='cuda:0', grad_fn=<NllLossBackward0>) 72200 tensor(4.8308, device='cuda:0', grad_fn=<NllLossBackward0>) 72300 tensor(4.7962, device='cuda:0', grad_fn=<NllLossBackward0>) 72400 tensor(4.9642, device='cuda:0', grad_fn=<NllLossBackward0>) 72500 tensor(4.3556, device='cuda:0', grad_fn=<NllLossBackward0>) 72600 tensor(5.0377, device='cuda:0', grad_fn=<NllLossBackward0>) 72700 tensor(5.0918, device='cuda:0', grad_fn=<NllLossBackward0>) 72800 tensor(4.6903, device='cuda:0', grad_fn=<NllLossBackward0>) 72900 tensor(4.8922, device='cuda:0', grad_fn=<NllLossBackward0>) 73000 tensor(4.6560, device='cuda:0', grad_fn=<NllLossBackward0>) 73100 tensor(4.5929, device='cuda:0', grad_fn=<NllLossBackward0>) 73200 tensor(4.9704, device='cuda:0', grad_fn=<NllLossBackward0>) 73300 tensor(4.6826, device='cuda:0', grad_fn=<NllLossBackward0>) 73400 tensor(4.8576, device='cuda:0', grad_fn=<NllLossBackward0>) 73500 tensor(4.5133, device='cuda:0', grad_fn=<NllLossBackward0>) 73600 tensor(5.0396, device='cuda:0', grad_fn=<NllLossBackward0>) 73700 tensor(4.7019, device='cuda:0', grad_fn=<NllLossBackward0>) 73800 tensor(4.6526, device='cuda:0', grad_fn=<NllLossBackward0>) 73900 tensor(4.7282, device='cuda:0', grad_fn=<NllLossBackward0>) 74000 tensor(4.8241, device='cuda:0', grad_fn=<NllLossBackward0>) 74100 tensor(4.8778, device='cuda:0', grad_fn=<NllLossBackward0>) 74200 tensor(5.0028, device='cuda:0', grad_fn=<NllLossBackward0>) 74300 tensor(4.9495, device='cuda:0', grad_fn=<NllLossBackward0>) 74400 tensor(5.0913, device='cuda:0', grad_fn=<NllLossBackward0>) 74500 tensor(5.1775, device='cuda:0', grad_fn=<NllLossBackward0>) 74600 tensor(4.4610, device='cuda:0', grad_fn=<NllLossBackward0>) 74700 tensor(5.3130, device='cuda:0', grad_fn=<NllLossBackward0>) 74800 tensor(4.7988, device='cuda:0', grad_fn=<NllLossBackward0>) 74900 tensor(4.7432, device='cuda:0', grad_fn=<NllLossBackward0>) 75000 tensor(5.1351, device='cuda:0', grad_fn=<NllLossBackward0>) 75100 tensor(4.7960, device='cuda:0', grad_fn=<NllLossBackward0>) 75200 tensor(4.6802, device='cuda:0', grad_fn=<NllLossBackward0>) 75300 tensor(4.9530, device='cuda:0', grad_fn=<NllLossBackward0>) 75400 tensor(5.1936, device='cuda:0', grad_fn=<NllLossBackward0>) 75500 tensor(4.9072, device='cuda:0', grad_fn=<NllLossBackward0>) 75600 tensor(4.9389, device='cuda:0', grad_fn=<NllLossBackward0>) 75700 tensor(4.7479, device='cuda:0', grad_fn=<NllLossBackward0>) 75800 tensor(5.0113, device='cuda:0', grad_fn=<NllLossBackward0>) 75900 tensor(4.8402, device='cuda:0', grad_fn=<NllLossBackward0>) 76000 tensor(4.7632, device='cuda:0', grad_fn=<NllLossBackward0>) 76100 tensor(5.0242, device='cuda:0', grad_fn=<NllLossBackward0>) 76200 tensor(5.1758, device='cuda:0', grad_fn=<NllLossBackward0>) 76300 tensor(5.2158, device='cuda:0', grad_fn=<NllLossBackward0>) 76400 tensor(4.7053, device='cuda:0', grad_fn=<NllLossBackward0>) 76500 tensor(4.6121, device='cuda:0', grad_fn=<NllLossBackward0>) 76600 tensor(4.8127, device='cuda:0', grad_fn=<NllLossBackward0>) 76700 tensor(4.7923, device='cuda:0', grad_fn=<NllLossBackward0>) 76800 tensor(4.9212, device='cuda:0', grad_fn=<NllLossBackward0>) 76900 tensor(4.7273, device='cuda:0', grad_fn=<NllLossBackward0>) 77000 tensor(5.0132, device='cuda:0', grad_fn=<NllLossBackward0>) 77100 tensor(5.0376, device='cuda:0', grad_fn=<NllLossBackward0>) 77200 tensor(5.1086, device='cuda:0', grad_fn=<NllLossBackward0>) 77300 tensor(5.0956, device='cuda:0', grad_fn=<NllLossBackward0>) 77400 tensor(4.2917, device='cuda:0', grad_fn=<NllLossBackward0>) 77500 tensor(4.7433, device='cuda:0', grad_fn=<NllLossBackward0>) 77600 tensor(5.1936, device='cuda:0', grad_fn=<NllLossBackward0>) 77700 tensor(4.6839, device='cuda:0', grad_fn=<NllLossBackward0>) 77800 tensor(4.9525, device='cuda:0', grad_fn=<NllLossBackward0>) 77900 tensor(4.9587, device='cuda:0', grad_fn=<NllLossBackward0>) 78000 tensor(4.6259, device='cuda:0', grad_fn=<NllLossBackward0>) 78100 tensor(5.1369, device='cuda:0', grad_fn=<NllLossBackward0>) 78200 tensor(4.8546, device='cuda:0', grad_fn=<NllLossBackward0>) 78300 tensor(4.7167, device='cuda:0', grad_fn=<NllLossBackward0>) 78400 tensor(4.8111, device='cuda:0', grad_fn=<NllLossBackward0>) 78500 tensor(5.0958, device='cuda:0', grad_fn=<NllLossBackward0>) 78600 tensor(4.6138, device='cuda:0', grad_fn=<NllLossBackward0>) 78700 tensor(4.8627, device='cuda:0', grad_fn=<NllLossBackward0>) 78800 tensor(4.7114, device='cuda:0', grad_fn=<NllLossBackward0>) 78900 tensor(4.9337, device='cuda:0', grad_fn=<NllLossBackward0>) 79000 tensor(4.8261, device='cuda:0', grad_fn=<NllLossBackward0>) 79100 tensor(4.8417, device='cuda:0', grad_fn=<NllLossBackward0>) 79200 tensor(4.9143, device='cuda:0', grad_fn=<NllLossBackward0>) 79300 tensor(5.1830, device='cuda:0', grad_fn=<NllLossBackward0>) 79400 tensor(4.7524, device='cuda:0', grad_fn=<NllLossBackward0>) 79500 tensor(4.8418, device='cuda:0', grad_fn=<NllLossBackward0>) 79600 tensor(4.8417, device='cuda:0', grad_fn=<NllLossBackward0>) 79700 tensor(4.8996, device='cuda:0', grad_fn=<NllLossBackward0>) 79800 tensor(5.0514, device='cuda:0', grad_fn=<NllLossBackward0>) 79900 tensor(4.7655, device='cuda:0', grad_fn=<NllLossBackward0>) 80000 tensor(5.0455, device='cuda:0', grad_fn=<NllLossBackward0>) 80100 tensor(4.7676, device='cuda:0', grad_fn=<NllLossBackward0>) 80200 tensor(4.9099, device='cuda:0', grad_fn=<NllLossBackward0>) 80300 tensor(4.6653, device='cuda:0', grad_fn=<NllLossBackward0>) 80400 tensor(5.4648, device='cuda:0', grad_fn=<NllLossBackward0>) 80500 tensor(5.2843, device='cuda:0', grad_fn=<NllLossBackward0>) 80600 tensor(4.9210, device='cuda:0', grad_fn=<NllLossBackward0>) 80700 tensor(5.0038, device='cuda:0', grad_fn=<NllLossBackward0>) 80800 tensor(4.9964, device='cuda:0', grad_fn=<NllLossBackward0>) 80900 tensor(4.6542, device='cuda:0', grad_fn=<NllLossBackward0>) 81000 tensor(4.4927, device='cuda:0', grad_fn=<NllLossBackward0>) 81100 tensor(4.9112, device='cuda:0', grad_fn=<NllLossBackward0>) 81200 tensor(4.7454, device='cuda:0', grad_fn=<NllLossBackward0>) 81300 tensor(4.7941, device='cuda:0', grad_fn=<NllLossBackward0>) 81400 tensor(4.7367, device='cuda:0', grad_fn=<NllLossBackward0>) 81500 tensor(4.4278, device='cuda:0', grad_fn=<NllLossBackward0>) 81600 tensor(4.6781, device='cuda:0', grad_fn=<NllLossBackward0>) 81700 tensor(4.7829, device='cuda:0', grad_fn=<NllLossBackward0>) 81800 tensor(5.2671, device='cuda:0', grad_fn=<NllLossBackward0>) 81900 tensor(5.0455, device='cuda:0', grad_fn=<NllLossBackward0>) 82000 tensor(4.5989, device='cuda:0', grad_fn=<NllLossBackward0>) 82100 tensor(4.8426, device='cuda:0', grad_fn=<NllLossBackward0>) 82200 tensor(5.1518, device='cuda:0', grad_fn=<NllLossBackward0>) 82300 tensor(4.9861, device='cuda:0', grad_fn=<NllLossBackward0>) 82400 tensor(4.7219, device='cuda:0', grad_fn=<NllLossBackward0>) 82500 tensor(4.4432, device='cuda:0', grad_fn=<NllLossBackward0>) 82600 tensor(5.1153, device='cuda:0', grad_fn=<NllLossBackward0>) 82700 tensor(5.0588, device='cuda:0', grad_fn=<NllLossBackward0>) 82800 tensor(4.9451, device='cuda:0', grad_fn=<NllLossBackward0>) 82900 tensor(4.8679, device='cuda:0', grad_fn=<NllLossBackward0>) 83000 tensor(4.9801, device='cuda:0', grad_fn=<NllLossBackward0>) 83100 tensor(4.5352, device='cuda:0', grad_fn=<NllLossBackward0>) 83200 tensor(5.1042, device='cuda:0', grad_fn=<NllLossBackward0>) 83300 tensor(4.8451, device='cuda:0', grad_fn=<NllLossBackward0>) 83400 tensor(5.1197, device='cuda:0', grad_fn=<NllLossBackward0>) 83500 tensor(5.0546, device='cuda:0', grad_fn=<NllLossBackward0>) 83600 tensor(4.8199, device='cuda:0', grad_fn=<NllLossBackward0>) 83700 tensor(5.0084, device='cuda:0', grad_fn=<NllLossBackward0>) 83800 tensor(4.8168, device='cuda:0', grad_fn=<NllLossBackward0>) 83900 tensor(4.8227, device='cuda:0', grad_fn=<NllLossBackward0>) 84000 tensor(4.7605, device='cuda:0', grad_fn=<NllLossBackward0>) 84100 tensor(4.5356, device='cuda:0', grad_fn=<NllLossBackward0>) 84200 tensor(4.4155, device='cuda:0', grad_fn=<NllLossBackward0>) 84300 tensor(5.0845, device='cuda:0', grad_fn=<NllLossBackward0>) 84400 tensor(5.0510, device='cuda:0', grad_fn=<NllLossBackward0>) 84500 tensor(4.9410, device='cuda:0', grad_fn=<NllLossBackward0>) 84600 tensor(4.8332, device='cuda:0', grad_fn=<NllLossBackward0>) 84700 tensor(4.7127, device='cuda:0', grad_fn=<NllLossBackward0>) 84800 tensor(4.8011, device='cuda:0', grad_fn=<NllLossBackward0>) 84900 tensor(4.7756, device='cuda:0', grad_fn=<NllLossBackward0>) 85000 tensor(4.6604, device='cuda:0', grad_fn=<NllLossBackward0>) 85100 tensor(4.5451, device='cuda:0', grad_fn=<NllLossBackward0>) 85200 tensor(4.8663, device='cuda:0', grad_fn=<NllLossBackward0>) 85300 tensor(5.0541, device='cuda:0', grad_fn=<NllLossBackward0>) 85400 tensor(5.0160, device='cuda:0', grad_fn=<NllLossBackward0>) 85500 tensor(5.1305, device='cuda:0', grad_fn=<NllLossBackward0>) 85600 tensor(4.7185, device='cuda:0', grad_fn=<NllLossBackward0>) 85700 tensor(4.1517, device='cuda:0', grad_fn=<NllLossBackward0>) 85800 tensor(5.3590, device='cuda:0', grad_fn=<NllLossBackward0>) 85900 tensor(4.8706, device='cuda:0', grad_fn=<NllLossBackward0>) 86000 tensor(4.9743, device='cuda:0', grad_fn=<NllLossBackward0>) 86100 tensor(4.9927, device='cuda:0', grad_fn=<NllLossBackward0>) 86200 tensor(4.9393, device='cuda:0', grad_fn=<NllLossBackward0>) 86300 tensor(4.6270, device='cuda:0', grad_fn=<NllLossBackward0>) 86400 tensor(5.1570, device='cuda:0', grad_fn=<NllLossBackward0>) 86500 tensor(4.9389, device='cuda:0', grad_fn=<NllLossBackward0>) 86600 tensor(4.8239, device='cuda:0', grad_fn=<NllLossBackward0>) 86700 tensor(4.9450, device='cuda:0', grad_fn=<NllLossBackward0>) 86800 tensor(4.6683, device='cuda:0', grad_fn=<NllLossBackward0>) 86900 tensor(4.9541, device='cuda:0', grad_fn=<NllLossBackward0>) 87000 tensor(4.9188, device='cuda:0', grad_fn=<NllLossBackward0>) 87100 tensor(4.9799, device='cuda:0', grad_fn=<NllLossBackward0>) 87200 tensor(5.0830, device='cuda:0', grad_fn=<NllLossBackward0>) 87300 tensor(5.0182, device='cuda:0', grad_fn=<NllLossBackward0>) 87400 tensor(5.0457, device='cuda:0', grad_fn=<NllLossBackward0>) 87500 tensor(5.0434, device='cuda:0', grad_fn=<NllLossBackward0>) 87600 tensor(5.0894, device='cuda:0', grad_fn=<NllLossBackward0>) 87700 tensor(4.8774, device='cuda:0', grad_fn=<NllLossBackward0>) 87800 tensor(4.6824, device='cuda:0', grad_fn=<NllLossBackward0>) 87900 tensor(5.0711, device='cuda:0', grad_fn=<NllLossBackward0>) 88000 tensor(4.6619, device='cuda:0', grad_fn=<NllLossBackward0>) 88100 tensor(4.7817, device='cuda:0', grad_fn=<NllLossBackward0>) 88200 tensor(4.6248, device='cuda:0', grad_fn=<NllLossBackward0>) 88300 tensor(4.6969, device='cuda:0', grad_fn=<NllLossBackward0>) 88400 tensor(5.1408, device='cuda:0', grad_fn=<NllLossBackward0>) 88500 tensor(4.9432, device='cuda:0', grad_fn=<NllLossBackward0>) 88600 tensor(5.0958, device='cuda:0', grad_fn=<NllLossBackward0>) 88700 tensor(4.9197, device='cuda:0', grad_fn=<NllLossBackward0>) 88800 tensor(4.8189, device='cuda:0', grad_fn=<NllLossBackward0>) 88900 tensor(5.0727, device='cuda:0', grad_fn=<NllLossBackward0>) 89000 tensor(4.7310, device='cuda:0', grad_fn=<NllLossBackward0>) 89100 tensor(4.8091, device='cuda:0', grad_fn=<NllLossBackward0>) 89200 tensor(4.9663, device='cuda:0', grad_fn=<NllLossBackward0>) 89300 tensor(4.8672, device='cuda:0', grad_fn=<NllLossBackward0>) 89400 tensor(4.7209, device='cuda:0', grad_fn=<NllLossBackward0>) 89500 tensor(5.0378, device='cuda:0', grad_fn=<NllLossBackward0>) 89600 tensor(4.8318, device='cuda:0', grad_fn=<NllLossBackward0>) 89700 tensor(4.8880, device='cuda:0', grad_fn=<NllLossBackward0>) 89800 tensor(5.1295, device='cuda:0', grad_fn=<NllLossBackward0>) 89900 tensor(4.7790, device='cuda:0', grad_fn=<NllLossBackward0>) 90000 tensor(5.1169, device='cuda:0', grad_fn=<NllLossBackward0>) 90100 tensor(4.8325, device='cuda:0', grad_fn=<NllLossBackward0>) 90200 tensor(4.9860, device='cuda:0', grad_fn=<NllLossBackward0>) 90300 tensor(4.8658, device='cuda:0', grad_fn=<NllLossBackward0>) 90400 tensor(4.9993, device='cuda:0', grad_fn=<NllLossBackward0>) 90500 tensor(4.6549, device='cuda:0', grad_fn=<NllLossBackward0>) 90600 tensor(4.6946, device='cuda:0', grad_fn=<NllLossBackward0>) 90700 tensor(4.9381, device='cuda:0', grad_fn=<NllLossBackward0>) 90800 tensor(4.7142, device='cuda:0', grad_fn=<NllLossBackward0>) 90900 tensor(4.7790, device='cuda:0', grad_fn=<NllLossBackward0>) 91000 tensor(5.2347, device='cuda:0', grad_fn=<NllLossBackward0>) 91100 tensor(4.8655, device='cuda:0', grad_fn=<NllLossBackward0>) 91200 tensor(4.5499, device='cuda:0', grad_fn=<NllLossBackward0>) 91300 tensor(4.8705, device='cuda:0', grad_fn=<NllLossBackward0>) 91400 tensor(4.7983, device='cuda:0', grad_fn=<NllLossBackward0>) 91500 tensor(4.5896, device='cuda:0', grad_fn=<NllLossBackward0>) 91600 tensor(4.8234, device='cuda:0', grad_fn=<NllLossBackward0>) 91700 tensor(4.9716, device='cuda:0', grad_fn=<NllLossBackward0>) 91800 tensor(4.7786, device='cuda:0', grad_fn=<NllLossBackward0>) 91900 tensor(4.7490, device='cuda:0', grad_fn=<NllLossBackward0>) 92000 tensor(5.2922, device='cuda:0', grad_fn=<NllLossBackward0>) 92100 tensor(5.0743, device='cuda:0', grad_fn=<NllLossBackward0>) 92200 tensor(4.9666, device='cuda:0', grad_fn=<NllLossBackward0>) 92300 tensor(4.9775, device='cuda:0', grad_fn=<NllLossBackward0>) 92400 tensor(4.9803, device='cuda:0', grad_fn=<NllLossBackward0>) 92500 tensor(4.7901, device='cuda:0', grad_fn=<NllLossBackward0>) 92600 tensor(4.7642, device='cuda:0', grad_fn=<NllLossBackward0>) 92700 tensor(4.8326, device='cuda:0', grad_fn=<NllLossBackward0>) 92800 tensor(4.9265, device='cuda:0', grad_fn=<NllLossBackward0>) 92900 tensor(4.5904, device='cuda:0', grad_fn=<NllLossBackward0>) 93000 tensor(5.0385, device='cuda:0', grad_fn=<NllLossBackward0>) 93100 tensor(4.4090, device='cuda:0', grad_fn=<NllLossBackward0>) 93200 tensor(4.8539, device='cuda:0', grad_fn=<NllLossBackward0>) 93300 tensor(4.8513, device='cuda:0', grad_fn=<NllLossBackward0>) 93400 tensor(5.1419, device='cuda:0', grad_fn=<NllLossBackward0>) 93500 tensor(4.7645, device='cuda:0', grad_fn=<NllLossBackward0>) 93600 tensor(4.9421, device='cuda:0', grad_fn=<NllLossBackward0>) 93700 tensor(5.0132, device='cuda:0', grad_fn=<NllLossBackward0>) 93800 tensor(4.7757, device='cuda:0', grad_fn=<NllLossBackward0>) 93900 tensor(4.8692, device='cuda:0', grad_fn=<NllLossBackward0>) 94000 tensor(4.5337, device='cuda:0', grad_fn=<NllLossBackward0>) 94100 tensor(4.9413, device='cuda:0', grad_fn=<NllLossBackward0>) 94200 tensor(4.6704, device='cuda:0', grad_fn=<NllLossBackward0>) 94300 tensor(4.7524, device='cuda:0', grad_fn=<NllLossBackward0>) 94400 tensor(4.7512, device='cuda:0', grad_fn=<NllLossBackward0>) 94500 tensor(4.7677, device='cuda:0', grad_fn=<NllLossBackward0>) 94600 tensor(4.7220, device='cuda:0', grad_fn=<NllLossBackward0>) 94700 tensor(5.1174, device='cuda:0', grad_fn=<NllLossBackward0>) 94800 tensor(4.7162, device='cuda:0', grad_fn=<NllLossBackward0>) 94900 tensor(4.8592, device='cuda:0', grad_fn=<NllLossBackward0>) 95000 tensor(5.1584, device='cuda:0', grad_fn=<NllLossBackward0>) 95100 tensor(4.6757, device='cuda:0', grad_fn=<NllLossBackward0>) 95200 tensor(4.9543, device='cuda:0', grad_fn=<NllLossBackward0>) 95300 tensor(4.8269, device='cuda:0', grad_fn=<NllLossBackward0>) 95400 tensor(4.3439, device='cuda:0', grad_fn=<NllLossBackward0>) 95500 tensor(4.7870, device='cuda:0', grad_fn=<NllLossBackward0>) 95600 tensor(4.9020, device='cuda:0', grad_fn=<NllLossBackward0>) 95700 tensor(4.5368, device='cuda:0', grad_fn=<NllLossBackward0>) 95800 tensor(5.0365, device='cuda:0', grad_fn=<NllLossBackward0>) 95900 tensor(5.1015, device='cuda:0', grad_fn=<NllLossBackward0>) 96000 tensor(5.1919, device='cuda:0', grad_fn=<NllLossBackward0>) 96100 tensor(5.0134, device='cuda:0', grad_fn=<NllLossBackward0>) 96200 tensor(4.9848, device='cuda:0', grad_fn=<NllLossBackward0>) 96300 tensor(4.8372, device='cuda:0', grad_fn=<NllLossBackward0>) 96400 tensor(5.0092, device='cuda:0', grad_fn=<NllLossBackward0>) 96500 tensor(4.9956, device='cuda:0', grad_fn=<NllLossBackward0>) 96600 tensor(4.8314, device='cuda:0', grad_fn=<NllLossBackward0>) 96700 tensor(4.8101, device='cuda:0', grad_fn=<NllLossBackward0>) 96800 tensor(4.9651, device='cuda:0', grad_fn=<NllLossBackward0>) 96900 tensor(4.8259, device='cuda:0', grad_fn=<NllLossBackward0>) 97000 tensor(4.7303, device='cuda:0', grad_fn=<NllLossBackward0>) 97100 tensor(5.0106, device='cuda:0', grad_fn=<NllLossBackward0>) 97200 tensor(4.6987, device='cuda:0', grad_fn=<NllLossBackward0>) 97300 tensor(4.7748, device='cuda:0', grad_fn=<NllLossBackward0>) 97400 tensor(4.8300, device='cuda:0', grad_fn=<NllLossBackward0>) 97500 tensor(4.9247, device='cuda:0', grad_fn=<NllLossBackward0>) 97600 tensor(4.7935, device='cuda:0', grad_fn=<NllLossBackward0>) 97700 tensor(4.6968, device='cuda:0', grad_fn=<NllLossBackward0>) 97800 tensor(4.9424, device='cuda:0', grad_fn=<NllLossBackward0>) 97900 tensor(4.5449, device='cuda:0', grad_fn=<NllLossBackward0>) 98000 tensor(4.8785, device='cuda:0', grad_fn=<NllLossBackward0>) 98100 tensor(5.2805, device='cuda:0', grad_fn=<NllLossBackward0>) 98200 tensor(5.0360, device='cuda:0', grad_fn=<NllLossBackward0>) 98300 tensor(4.8959, device='cuda:0', grad_fn=<NllLossBackward0>) 98400 tensor(4.8205, device='cuda:0', grad_fn=<NllLossBackward0>) 98500 tensor(4.8109, device='cuda:0', grad_fn=<NllLossBackward0>) 98600 tensor(4.7517, device='cuda:0', grad_fn=<NllLossBackward0>) 98700 tensor(4.2260, device='cuda:0', grad_fn=<NllLossBackward0>) 98800 tensor(4.8782, device='cuda:0', grad_fn=<NllLossBackward0>) 98900 tensor(4.7034, device='cuda:0', grad_fn=<NllLossBackward0>) 99000 tensor(4.5861, device='cuda:0', grad_fn=<NllLossBackward0>) 99100 tensor(4.9904, device='cuda:0', grad_fn=<NllLossBackward0>) 99200 tensor(4.9637, device='cuda:0', grad_fn=<NllLossBackward0>) 99300 tensor(4.7789, device='cuda:0', grad_fn=<NllLossBackward0>) 99400 tensor(4.9661, device='cuda:0', grad_fn=<NllLossBackward0>) 99500 tensor(4.4535, device='cuda:0', grad_fn=<NllLossBackward0>) 99600 tensor(4.7522, device='cuda:0', grad_fn=<NllLossBackward0>) 99700 tensor(4.1814, device='cuda:0', grad_fn=<NllLossBackward0>) 99800 tensor(4.4184, device='cuda:0', grad_fn=<NllLossBackward0>) 99900 tensor(4.9469, device='cuda:0', grad_fn=<NllLossBackward0>) 100000 tensor(4.9216, device='cuda:0', grad_fn=<NllLossBackward0>) 100100 tensor(5.0670, device='cuda:0', grad_fn=<NllLossBackward0>) 100200 tensor(4.4815, device='cuda:0', grad_fn=<NllLossBackward0>) 100300 tensor(4.8047, device='cuda:0', grad_fn=<NllLossBackward0>) 100400 tensor(4.7661, device='cuda:0', grad_fn=<NllLossBackward0>) 100500 tensor(5.0412, device='cuda:0', grad_fn=<NllLossBackward0>) 100600 tensor(5.0822, device='cuda:0', grad_fn=<NllLossBackward0>) 100700 tensor(4.9617, device='cuda:0', grad_fn=<NllLossBackward0>) 100800 tensor(5.0497, device='cuda:0', grad_fn=<NllLossBackward0>) 100900 tensor(5.0292, device='cuda:0', grad_fn=<NllLossBackward0>) 101000 tensor(4.3783, device='cuda:0', grad_fn=<NllLossBackward0>) 101100 tensor(5.1925, device='cuda:0', grad_fn=<NllLossBackward0>) 101200 tensor(4.8833, device='cuda:0', grad_fn=<NllLossBackward0>) 101300 tensor(4.8256, device='cuda:0', grad_fn=<NllLossBackward0>) 101400 tensor(5.1346, device='cuda:0', grad_fn=<NllLossBackward0>) 101500 tensor(4.9421, device='cuda:0', grad_fn=<NllLossBackward0>) 101600 tensor(4.7890, device='cuda:0', grad_fn=<NllLossBackward0>) 101700 tensor(4.5993, device='cuda:0', grad_fn=<NllLossBackward0>) 101800 tensor(4.8848, device='cuda:0', grad_fn=<NllLossBackward0>) 101900 tensor(5.0264, device='cuda:0', grad_fn=<NllLossBackward0>) 102000 tensor(4.5711, device='cuda:0', grad_fn=<NllLossBackward0>) 102100 tensor(5.0150, device='cuda:0', grad_fn=<NllLossBackward0>) 102200 tensor(4.8943, device='cuda:0', grad_fn=<NllLossBackward0>) 102300 tensor(4.8837, device='cuda:0', grad_fn=<NllLossBackward0>) 102400 tensor(5.0701, device='cuda:0', grad_fn=<NllLossBackward0>) 102500 tensor(4.6735, device='cuda:0', grad_fn=<NllLossBackward0>) 102600 tensor(4.7247, device='cuda:0', grad_fn=<NllLossBackward0>) 102700 tensor(4.7052, device='cuda:0', grad_fn=<NllLossBackward0>) 102800 tensor(4.4751, device='cuda:0', grad_fn=<NllLossBackward0>) 102900 tensor(4.3246, device='cuda:0', grad_fn=<NllLossBackward0>) 103000 tensor(4.8290, device='cuda:0', grad_fn=<NllLossBackward0>) 103100 tensor(4.8880, device='cuda:0', grad_fn=<NllLossBackward0>) 103200 tensor(4.7565, device='cuda:0', grad_fn=<NllLossBackward0>) 103300 tensor(4.8007, device='cuda:0', grad_fn=<NllLossBackward0>) 103400 tensor(4.9262, device='cuda:0', grad_fn=<NllLossBackward0>) 103500 tensor(4.9762, device='cuda:0', grad_fn=<NllLossBackward0>) 103600 tensor(4.9669, device='cuda:0', grad_fn=<NllLossBackward0>) 103700 tensor(5.1680, device='cuda:0', grad_fn=<NllLossBackward0>) 103800 tensor(5.0296, device='cuda:0', grad_fn=<NllLossBackward0>) 103900 tensor(4.9457, device='cuda:0', grad_fn=<NllLossBackward0>) 104000 tensor(4.9833, device='cuda:0', grad_fn=<NllLossBackward0>) 104100 tensor(5.1114, device='cuda:0', grad_fn=<NllLossBackward0>) 104200 tensor(4.8435, device='cuda:0', grad_fn=<NllLossBackward0>) 104300 tensor(4.5540, device='cuda:0', grad_fn=<NllLossBackward0>) 104400 tensor(4.5313, device='cuda:0', grad_fn=<NllLossBackward0>) 104500 tensor(5.0209, device='cuda:0', grad_fn=<NllLossBackward0>) 104600 tensor(4.1863, device='cuda:0', grad_fn=<NllLossBackward0>) 104700 tensor(4.9792, device='cuda:0', grad_fn=<NllLossBackward0>) 104800 tensor(4.7222, device='cuda:0', grad_fn=<NllLossBackward0>) 104900 tensor(4.8395, device='cuda:0', grad_fn=<NllLossBackward0>) 105000 tensor(4.8403, device='cuda:0', grad_fn=<NllLossBackward0>) 105100 tensor(4.9299, device='cuda:0', grad_fn=<NllLossBackward0>) 105200 tensor(4.9324, device='cuda:0', grad_fn=<NllLossBackward0>) 105300 tensor(5.0470, device='cuda:0', grad_fn=<NllLossBackward0>) 105400 tensor(4.7876, device='cuda:0', grad_fn=<NllLossBackward0>) 105500 tensor(5.0069, device='cuda:0', grad_fn=<NllLossBackward0>) 105600 tensor(5.0764, device='cuda:0', grad_fn=<NllLossBackward0>) 105700 tensor(5.2300, device='cuda:0', grad_fn=<NllLossBackward0>) 105800 tensor(4.8332, device='cuda:0', grad_fn=<NllLossBackward0>) 105900 tensor(5.0333, device='cuda:0', grad_fn=<NllLossBackward0>) 106000 tensor(4.9133, device='cuda:0', grad_fn=<NllLossBackward0>) 106100 tensor(4.8769, device='cuda:0', grad_fn=<NllLossBackward0>) 106200 tensor(5.1212, device='cuda:0', grad_fn=<NllLossBackward0>) 106300 tensor(4.8909, device='cuda:0', grad_fn=<NllLossBackward0>) 106400 tensor(4.7504, device='cuda:0', grad_fn=<NllLossBackward0>) 106500 tensor(4.9875, device='cuda:0', grad_fn=<NllLossBackward0>) 106600 tensor(4.6188, device='cuda:0', grad_fn=<NllLossBackward0>) 106700 tensor(4.4476, device='cuda:0', grad_fn=<NllLossBackward0>) 106800 tensor(4.9015, device='cuda:0', grad_fn=<NllLossBackward0>) 106900 tensor(4.9961, device='cuda:0', grad_fn=<NllLossBackward0>) 107000 tensor(4.5983, device='cuda:0', grad_fn=<NllLossBackward0>) 107100 tensor(4.9465, device='cuda:0', grad_fn=<NllLossBackward0>) 107200 tensor(4.9934, device='cuda:0', grad_fn=<NllLossBackward0>) 107300 tensor(4.2177, device='cuda:0', grad_fn=<NllLossBackward0>) 107400 tensor(4.8564, device='cuda:0', grad_fn=<NllLossBackward0>) 107500 tensor(4.7367, device='cuda:0', grad_fn=<NllLossBackward0>) 107600 tensor(4.6914, device='cuda:0', grad_fn=<NllLossBackward0>) 107700 tensor(4.9439, device='cuda:0', grad_fn=<NllLossBackward0>) 107800 tensor(4.8153, device='cuda:0', grad_fn=<NllLossBackward0>) 107900 tensor(4.3782, device='cuda:0', grad_fn=<NllLossBackward0>) 108000 tensor(5.0639, device='cuda:0', grad_fn=<NllLossBackward0>) 108100 tensor(5.0140, device='cuda:0', grad_fn=<NllLossBackward0>) 108200 tensor(4.8026, device='cuda:0', grad_fn=<NllLossBackward0>) 108300 tensor(5.0365, device='cuda:0', grad_fn=<NllLossBackward0>) 108400 tensor(5.0422, device='cuda:0', grad_fn=<NllLossBackward0>) 108500 tensor(4.7082, device='cuda:0', grad_fn=<NllLossBackward0>) 108600 tensor(4.5527, device='cuda:0', grad_fn=<NllLossBackward0>) 108700 tensor(4.7589, device='cuda:0', grad_fn=<NllLossBackward0>) 108800 tensor(5.0291, device='cuda:0', grad_fn=<NllLossBackward0>) 108900 tensor(4.5464, device='cuda:0', grad_fn=<NllLossBackward0>) 109000 tensor(4.9586, device='cuda:0', grad_fn=<NllLossBackward0>) 109100 tensor(4.9770, device='cuda:0', grad_fn=<NllLossBackward0>) 109200 tensor(4.9075, device='cuda:0', grad_fn=<NllLossBackward0>) 109300 tensor(4.5973, device='cuda:0', grad_fn=<NllLossBackward0>) 109400 tensor(5.0076, device='cuda:0', grad_fn=<NllLossBackward0>) 109500 tensor(4.7639, device='cuda:0', grad_fn=<NllLossBackward0>) 109600 tensor(4.7529, device='cuda:0', grad_fn=<NllLossBackward0>) 109700 tensor(5.0050, device='cuda:0', grad_fn=<NllLossBackward0>) 109800 tensor(5.0360, device='cuda:0', grad_fn=<NllLossBackward0>) 109900 tensor(4.7775, device='cuda:0', grad_fn=<NllLossBackward0>) 110000 tensor(4.6627, device='cuda:0', grad_fn=<NllLossBackward0>) 110100 tensor(4.8799, device='cuda:0', grad_fn=<NllLossBackward0>) 110200 tensor(4.7465, device='cuda:0', grad_fn=<NllLossBackward0>) 110300 tensor(5.0446, device='cuda:0', grad_fn=<NllLossBackward0>) 110400 tensor(4.8966, device='cuda:0', grad_fn=<NllLossBackward0>) 110500 tensor(4.9575, device='cuda:0', grad_fn=<NllLossBackward0>) 110600 tensor(4.8801, device='cuda:0', grad_fn=<NllLossBackward0>) 110700 tensor(4.9705, device='cuda:0', grad_fn=<NllLossBackward0>) 110800 tensor(4.8122, device='cuda:0', grad_fn=<NllLossBackward0>) 110900 tensor(4.7103, device='cuda:0', grad_fn=<NllLossBackward0>) 111000 tensor(4.7193, device='cuda:0', grad_fn=<NllLossBackward0>) 111100 tensor(4.9327, device='cuda:0', grad_fn=<NllLossBackward0>) 111200 tensor(4.9441, device='cuda:0', grad_fn=<NllLossBackward0>) 111300 tensor(5.2106, device='cuda:0', grad_fn=<NllLossBackward0>) 111400 tensor(4.5818, device='cuda:0', grad_fn=<NllLossBackward0>) 111500 tensor(4.6068, device='cuda:0', grad_fn=<NllLossBackward0>) 111600 tensor(5.0798, device='cuda:0', grad_fn=<NllLossBackward0>) 111700 tensor(5.1382, device='cuda:0', grad_fn=<NllLossBackward0>) 111800 tensor(4.3443, device='cuda:0', grad_fn=<NllLossBackward0>) 111900 tensor(5.0532, device='cuda:0', grad_fn=<NllLossBackward0>) 112000 tensor(4.6804, device='cuda:0', grad_fn=<NllLossBackward0>) 112100 tensor(4.5050, device='cuda:0', grad_fn=<NllLossBackward0>) 112200 tensor(4.9678, device='cuda:0', grad_fn=<NllLossBackward0>) 112300 tensor(5.1017, device='cuda:0', grad_fn=<NllLossBackward0>) 112400 tensor(4.9474, device='cuda:0', grad_fn=<NllLossBackward0>) 112500 tensor(5.0659, device='cuda:0', grad_fn=<NllLossBackward0>) 112600 tensor(4.1369, device='cuda:0', grad_fn=<NllLossBackward0>) 112700 tensor(4.8935, device='cuda:0', grad_fn=<NllLossBackward0>) 112800 tensor(4.6240, device='cuda:0', grad_fn=<NllLossBackward0>) 112900 tensor(4.8367, device='cuda:0', grad_fn=<NllLossBackward0>) 113000 tensor(4.6743, device='cuda:0', grad_fn=<NllLossBackward0>) 113100 tensor(5.0598, device='cuda:0', grad_fn=<NllLossBackward0>) 113200 tensor(4.7863, device='cuda:0', grad_fn=<NllLossBackward0>) 113300 tensor(3.9544, device='cuda:0', grad_fn=<NllLossBackward0>) 113400 tensor(5.0014, device='cuda:0', grad_fn=<NllLossBackward0>) 113500 tensor(4.8981, device='cuda:0', grad_fn=<NllLossBackward0>) 113600 tensor(4.6241, device='cuda:0', grad_fn=<NllLossBackward0>) 113700 tensor(4.7643, device='cuda:0', grad_fn=<NllLossBackward0>) 113800 tensor(4.4898, device='cuda:0', grad_fn=<NllLossBackward0>) 113900 tensor(4.7177, device='cuda:0', grad_fn=<NllLossBackward0>) 114000 tensor(5.1589, device='cuda:0', grad_fn=<NllLossBackward0>) 114100 tensor(4.8498, device='cuda:0', grad_fn=<NllLossBackward0>) 114200 tensor(4.6020, device='cuda:0', grad_fn=<NllLossBackward0>) 114300 tensor(4.8476, device='cuda:0', grad_fn=<NllLossBackward0>) 114400 tensor(4.9836, device='cuda:0', grad_fn=<NllLossBackward0>) 114500 tensor(5.2996, device='cuda:0', grad_fn=<NllLossBackward0>) 114600 tensor(5.2211, device='cuda:0', grad_fn=<NllLossBackward0>) 114700 tensor(4.6157, device='cuda:0', grad_fn=<NllLossBackward0>) 114800 tensor(4.9978, device='cuda:0', grad_fn=<NllLossBackward0>) 114900 tensor(4.7124, device='cuda:0', grad_fn=<NllLossBackward0>) 115000 tensor(4.8855, device='cuda:0', grad_fn=<NllLossBackward0>) 115100 tensor(4.8061, device='cuda:0', grad_fn=<NllLossBackward0>) 115200 tensor(4.9216, device='cuda:0', grad_fn=<NllLossBackward0>) 115300 tensor(4.7439, device='cuda:0', grad_fn=<NllLossBackward0>) 115400 tensor(5.1855, device='cuda:0', grad_fn=<NllLossBackward0>) 115500 tensor(5.0700, device='cuda:0', grad_fn=<NllLossBackward0>) 115600 tensor(4.9031, device='cuda:0', grad_fn=<NllLossBackward0>) 115700 tensor(4.9650, device='cuda:0', grad_fn=<NllLossBackward0>) 115800 tensor(4.7938, device='cuda:0', grad_fn=<NllLossBackward0>) 115900 tensor(4.9974, device='cuda:0', grad_fn=<NllLossBackward0>) 116000 tensor(4.7874, device='cuda:0', grad_fn=<NllLossBackward0>) 116100 tensor(4.8274, device='cuda:0', grad_fn=<NllLossBackward0>) 116200 tensor(4.8384, device='cuda:0', grad_fn=<NllLossBackward0>) 116300 tensor(5.1083, device='cuda:0', grad_fn=<NllLossBackward0>) 116400 tensor(4.6231, device='cuda:0', grad_fn=<NllLossBackward0>) 116500 tensor(4.5965, device='cuda:0', grad_fn=<NllLossBackward0>) 116600 tensor(4.6714, device='cuda:0', grad_fn=<NllLossBackward0>) 116700 tensor(4.9358, device='cuda:0', grad_fn=<NllLossBackward0>) 116800 tensor(5.0778, device='cuda:0', grad_fn=<NllLossBackward0>) 116900 tensor(4.7613, device='cuda:0', grad_fn=<NllLossBackward0>) 117000 tensor(4.6919, device='cuda:0', grad_fn=<NllLossBackward0>) 117100 tensor(4.7312, device='cuda:0', grad_fn=<NllLossBackward0>) 117200 tensor(4.8210, device='cuda:0', grad_fn=<NllLossBackward0>) 117300 tensor(4.3730, device='cuda:0', grad_fn=<NllLossBackward0>) 117400 tensor(4.8292, device='cuda:0', grad_fn=<NllLossBackward0>) 117500 tensor(4.7531, device='cuda:0', grad_fn=<NllLossBackward0>) 117600 tensor(4.8388, device='cuda:0', grad_fn=<NllLossBackward0>) 117700 tensor(4.8566, device='cuda:0', grad_fn=<NllLossBackward0>) 117800 tensor(4.9021, device='cuda:0', grad_fn=<NllLossBackward0>) 117900 tensor(4.8977, device='cuda:0', grad_fn=<NllLossBackward0>) 118000 tensor(4.7841, device='cuda:0', grad_fn=<NllLossBackward0>) 118100 tensor(4.6447, device='cuda:0', grad_fn=<NllLossBackward0>) 118200 tensor(4.5021, device='cuda:0', grad_fn=<NllLossBackward0>) 118300 tensor(4.9625, device='cuda:0', grad_fn=<NllLossBackward0>) 118400 tensor(5.0818, device='cuda:0', grad_fn=<NllLossBackward0>) 118500 tensor(4.9943, device='cuda:0', grad_fn=<NllLossBackward0>) 118600 tensor(4.8726, device='cuda:0', grad_fn=<NllLossBackward0>) 118700 tensor(4.9113, device='cuda:0', grad_fn=<NllLossBackward0>) 118800 tensor(4.5782, device='cuda:0', grad_fn=<NllLossBackward0>) 118900 tensor(4.4148, device='cuda:0', grad_fn=<NllLossBackward0>) 119000 tensor(5.0278, device='cuda:0', grad_fn=<NllLossBackward0>) 119100 tensor(5.2606, device='cuda:0', grad_fn=<NllLossBackward0>) 119200 tensor(4.9693, device='cuda:0', grad_fn=<NllLossBackward0>) 119300 tensor(4.6617, device='cuda:0', grad_fn=<NllLossBackward0>) 119400 tensor(4.8571, device='cuda:0', grad_fn=<NllLossBackward0>) 119500 tensor(4.2560, device='cuda:0', grad_fn=<NllLossBackward0>) 119600 tensor(4.9186, device='cuda:0', grad_fn=<NllLossBackward0>) 119700 tensor(4.9237, device='cuda:0', grad_fn=<NllLossBackward0>) 119800 tensor(4.9456, device='cuda:0', grad_fn=<NllLossBackward0>) 119900 tensor(4.9364, device='cuda:0', grad_fn=<NllLossBackward0>) 120000 tensor(4.5929, device='cuda:0', grad_fn=<NllLossBackward0>) 120100 tensor(4.9357, device='cuda:0', grad_fn=<NllLossBackward0>) 120200 tensor(4.8705, device='cuda:0', grad_fn=<NllLossBackward0>) 120300 tensor(4.7914, device='cuda:0', grad_fn=<NllLossBackward0>) 120400 tensor(4.7161, device='cuda:0', grad_fn=<NllLossBackward0>) 120500 tensor(4.7885, device='cuda:0', grad_fn=<NllLossBackward0>) 120600 tensor(4.8593, device='cuda:0', grad_fn=<NllLossBackward0>) 120700 tensor(4.7298, device='cuda:0', grad_fn=<NllLossBackward0>) 120800 tensor(4.8347, device='cuda:0', grad_fn=<NllLossBackward0>) 120900 tensor(4.9481, device='cuda:0', grad_fn=<NllLossBackward0>) 121000 tensor(4.6276, device='cuda:0', grad_fn=<NllLossBackward0>) 121100 tensor(4.6297, device='cuda:0', grad_fn=<NllLossBackward0>) 121200 tensor(4.4420, device='cuda:0', grad_fn=<NllLossBackward0>) 121300 tensor(4.9996, device='cuda:0', grad_fn=<NllLossBackward0>) 121400 tensor(4.6057, device='cuda:0', grad_fn=<NllLossBackward0>) 121500 tensor(5.2147, device='cuda:0', grad_fn=<NllLossBackward0>) 121600 tensor(5.0158, device='cuda:0', grad_fn=<NllLossBackward0>) 121700 tensor(4.7963, device='cuda:0', grad_fn=<NllLossBackward0>) 121800 tensor(5.0160, device='cuda:0', grad_fn=<NllLossBackward0>) 121900 tensor(4.9921, device='cuda:0', grad_fn=<NllLossBackward0>) 122000 tensor(4.9018, device='cuda:0', grad_fn=<NllLossBackward0>) 122100 tensor(4.9365, device='cuda:0', grad_fn=<NllLossBackward0>) 122200 tensor(4.6702, device='cuda:0', grad_fn=<NllLossBackward0>) 122300 tensor(4.9160, device='cuda:0', grad_fn=<NllLossBackward0>) 122400 tensor(4.9476, device='cuda:0', grad_fn=<NllLossBackward0>) 122500 tensor(4.8492, device='cuda:0', grad_fn=<NllLossBackward0>) 122600 tensor(4.6968, device='cuda:0', grad_fn=<NllLossBackward0>) 122700 tensor(4.7005, device='cuda:0', grad_fn=<NllLossBackward0>) 122800 tensor(4.4889, device='cuda:0', grad_fn=<NllLossBackward0>) 122900 tensor(4.4938, device='cuda:0', grad_fn=<NllLossBackward0>) 123000 tensor(4.8215, device='cuda:0', grad_fn=<NllLossBackward0>) 123100 tensor(4.6476, device='cuda:0', grad_fn=<NllLossBackward0>) 123200 tensor(4.6063, device='cuda:0', grad_fn=<NllLossBackward0>) 123300 tensor(4.7627, device='cuda:0', grad_fn=<NllLossBackward0>) 123400 tensor(4.9589, device='cuda:0', grad_fn=<NllLossBackward0>) 123500 tensor(5.0093, device='cuda:0', grad_fn=<NllLossBackward0>) 123600 tensor(5.1528, device='cuda:0', grad_fn=<NllLossBackward0>) 123700 tensor(4.5435, device='cuda:0', grad_fn=<NllLossBackward0>) 123800 tensor(5.0693, device='cuda:0', grad_fn=<NllLossBackward0>) 123900 tensor(4.4490, device='cuda:0', grad_fn=<NllLossBackward0>) 124000 tensor(4.9851, device='cuda:0', grad_fn=<NllLossBackward0>) 124100 tensor(4.9539, device='cuda:0', grad_fn=<NllLossBackward0>) 124200 tensor(5.0405, device='cuda:0', grad_fn=<NllLossBackward0>) 124300 tensor(4.6855, device='cuda:0', grad_fn=<NllLossBackward0>) 124400 tensor(4.9417, device='cuda:0', grad_fn=<NllLossBackward0>) 124500 tensor(4.4380, device='cuda:0', grad_fn=<NllLossBackward0>) 124600 tensor(4.0469, device='cuda:0', grad_fn=<NllLossBackward0>) 124700 tensor(5.0332, device='cuda:0', grad_fn=<NllLossBackward0>) 124800 tensor(4.9670, device='cuda:0', grad_fn=<NllLossBackward0>) 124900 tensor(5.0651, device='cuda:0', grad_fn=<NllLossBackward0>) 125000 tensor(4.5116, device='cuda:0', grad_fn=<NllLossBackward0>) 125100 tensor(5.0326, device='cuda:0', grad_fn=<NllLossBackward0>) 125200 tensor(4.7236, device='cuda:0', grad_fn=<NllLossBackward0>) 125300 tensor(4.7683, device='cuda:0', grad_fn=<NllLossBackward0>) 125400 tensor(4.5482, device='cuda:0', grad_fn=<NllLossBackward0>) 125500 tensor(4.5634, device='cuda:0', grad_fn=<NllLossBackward0>) 125600 tensor(4.4635, device='cuda:0', grad_fn=<NllLossBackward0>) 125700 tensor(4.4219, device='cuda:0', grad_fn=<NllLossBackward0>) 125800 tensor(4.9547, device='cuda:0', grad_fn=<NllLossBackward0>) 125900 tensor(4.8685, device='cuda:0', grad_fn=<NllLossBackward0>) 126000 tensor(4.3485, device='cuda:0', grad_fn=<NllLossBackward0>) 126100 tensor(4.5584, device='cuda:0', grad_fn=<NllLossBackward0>) 126200 tensor(4.6830, device='cuda:0', grad_fn=<NllLossBackward0>) 126300 tensor(5.0172, device='cuda:0', grad_fn=<NllLossBackward0>) 126400 tensor(4.9179, device='cuda:0', grad_fn=<NllLossBackward0>) 126500 tensor(4.8389, device='cuda:0', grad_fn=<NllLossBackward0>) 126600 tensor(4.7898, device='cuda:0', grad_fn=<NllLossBackward0>) 126700 tensor(4.9195, device='cuda:0', grad_fn=<NllLossBackward0>) 126800 tensor(4.8269, device='cuda:0', grad_fn=<NllLossBackward0>) 126900 tensor(4.8297, device='cuda:0', grad_fn=<NllLossBackward0>) 127000 tensor(4.8922, device='cuda:0', grad_fn=<NllLossBackward0>) 127100 tensor(4.8160, device='cuda:0', grad_fn=<NllLossBackward0>) 127200 tensor(4.7967, device='cuda:0', grad_fn=<NllLossBackward0>) 127300 tensor(4.7824, device='cuda:0', grad_fn=<NllLossBackward0>) 127400 tensor(4.8233, device='cuda:0', grad_fn=<NllLossBackward0>) 127500 tensor(4.9216, device='cuda:0', grad_fn=<NllLossBackward0>) 127600 tensor(5.0369, device='cuda:0', grad_fn=<NllLossBackward0>) 127700 tensor(4.8331, device='cuda:0', grad_fn=<NllLossBackward0>) 127800 tensor(4.8210, device='cuda:0', grad_fn=<NllLossBackward0>) 127900 tensor(4.8657, device='cuda:0', grad_fn=<NllLossBackward0>) 128000 tensor(4.7904, device='cuda:0', grad_fn=<NllLossBackward0>) 128100 tensor(4.6721, device='cuda:0', grad_fn=<NllLossBackward0>) 128200 tensor(5.0386, device='cuda:0', grad_fn=<NllLossBackward0>) 128300 tensor(4.5596, device='cuda:0', grad_fn=<NllLossBackward0>) 128400 tensor(4.5702, device='cuda:0', grad_fn=<NllLossBackward0>) 128500 tensor(4.7675, device='cuda:0', grad_fn=<NllLossBackward0>) 128600 tensor(4.8381, device='cuda:0', grad_fn=<NllLossBackward0>) 128700 tensor(4.5835, device='cuda:0', grad_fn=<NllLossBackward0>) 128800 tensor(4.7954, device='cuda:0', grad_fn=<NllLossBackward0>) 128900 tensor(4.7430, device='cuda:0', grad_fn=<NllLossBackward0>) 129000 tensor(4.6958, device='cuda:0', grad_fn=<NllLossBackward0>) 129100 tensor(4.4172, device='cuda:0', grad_fn=<NllLossBackward0>) 129200 tensor(4.8874, device='cuda:0', grad_fn=<NllLossBackward0>) 129300 tensor(4.3820, device='cuda:0', grad_fn=<NllLossBackward0>) 129400 tensor(4.8125, device='cuda:0', grad_fn=<NllLossBackward0>) 129500 tensor(4.9783, device='cuda:0', grad_fn=<NllLossBackward0>) 129600 tensor(4.5286, device='cuda:0', grad_fn=<NllLossBackward0>) 129700 tensor(4.8962, device='cuda:0', grad_fn=<NllLossBackward0>) 129800 tensor(4.6818, device='cuda:0', grad_fn=<NllLossBackward0>) 129900 tensor(5.0567, device='cuda:0', grad_fn=<NllLossBackward0>) 130000 tensor(5.1978, device='cuda:0', grad_fn=<NllLossBackward0>) 130100 tensor(5.1098, device='cuda:0', grad_fn=<NllLossBackward0>) 130200 tensor(4.6280, device='cuda:0', grad_fn=<NllLossBackward0>) 130300 tensor(4.9298, device='cuda:0', grad_fn=<NllLossBackward0>) 130400 tensor(4.6004, device='cuda:0', grad_fn=<NllLossBackward0>) 130500 tensor(4.8462, device='cuda:0', grad_fn=<NllLossBackward0>) 130600 tensor(4.7324, device='cuda:0', grad_fn=<NllLossBackward0>) 130700 tensor(4.7167, device='cuda:0', grad_fn=<NllLossBackward0>) 130800 tensor(4.4547, device='cuda:0', grad_fn=<NllLossBackward0>) 130900 tensor(4.5560, device='cuda:0', grad_fn=<NllLossBackward0>) 131000 tensor(4.0357, device='cuda:0', grad_fn=<NllLossBackward0>) 131100 tensor(4.8499, device='cuda:0', grad_fn=<NllLossBackward0>) 131200 tensor(4.9097, device='cuda:0', grad_fn=<NllLossBackward0>) 131300 tensor(4.8725, device='cuda:0', grad_fn=<NllLossBackward0>) 131400 tensor(4.8507, device='cuda:0', grad_fn=<NllLossBackward0>) 131500 tensor(4.8822, device='cuda:0', grad_fn=<NllLossBackward0>) 131600 tensor(4.8053, device='cuda:0', grad_fn=<NllLossBackward0>) 131700 tensor(4.7774, device='cuda:0', grad_fn=<NllLossBackward0>) 131800 tensor(4.6166, device='cuda:0', grad_fn=<NllLossBackward0>) 131900 tensor(5.0381, device='cuda:0', grad_fn=<NllLossBackward0>) 132000 tensor(5.1435, device='cuda:0', grad_fn=<NllLossBackward0>) 132100 tensor(4.9651, device='cuda:0', grad_fn=<NllLossBackward0>) 132200 tensor(4.3226, device='cuda:0', grad_fn=<NllLossBackward0>) 132300 tensor(4.9645, device='cuda:0', grad_fn=<NllLossBackward0>) 132400 tensor(5.1297, device='cuda:0', grad_fn=<NllLossBackward0>) 132500 tensor(4.9514, device='cuda:0', grad_fn=<NllLossBackward0>) 132600 tensor(4.4122, device='cuda:0', grad_fn=<NllLossBackward0>) 132700 tensor(4.8552, device='cuda:0', grad_fn=<NllLossBackward0>) 132800 tensor(4.4157, device='cuda:0', grad_fn=<NllLossBackward0>) 132900 tensor(4.9525, device='cuda:0', grad_fn=<NllLossBackward0>) 133000 tensor(4.7483, device='cuda:0', grad_fn=<NllLossBackward0>) 133100 tensor(4.6578, device='cuda:0', grad_fn=<NllLossBackward0>) 133200 tensor(5.0746, device='cuda:0', grad_fn=<NllLossBackward0>) 133300 tensor(4.8121, device='cuda:0', grad_fn=<NllLossBackward0>) 133400 tensor(4.8101, device='cuda:0', grad_fn=<NllLossBackward0>) 133500 tensor(4.6605, device='cuda:0', grad_fn=<NllLossBackward0>) 133600 tensor(4.7754, device='cuda:0', grad_fn=<NllLossBackward0>) 133700 tensor(4.4397, device='cuda:0', grad_fn=<NllLossBackward0>) 133800 tensor(4.8784, device='cuda:0', grad_fn=<NllLossBackward0>) 133900 tensor(4.6728, device='cuda:0', grad_fn=<NllLossBackward0>) 134000 tensor(4.7126, device='cuda:0', grad_fn=<NllLossBackward0>) 134100 tensor(4.5749, device='cuda:0', grad_fn=<NllLossBackward0>) 134200 tensor(4.4029, device='cuda:0', grad_fn=<NllLossBackward0>) 134300 tensor(4.7075, device='cuda:0', grad_fn=<NllLossBackward0>) 134400 tensor(5.0396, device='cuda:0', grad_fn=<NllLossBackward0>) 134500 tensor(4.6848, device='cuda:0', grad_fn=<NllLossBackward0>) 134600 tensor(5.0197, device='cuda:0', grad_fn=<NllLossBackward0>) 134700 tensor(4.7742, device='cuda:0', grad_fn=<NllLossBackward0>) 134800 tensor(4.8104, device='cuda:0', grad_fn=<NllLossBackward0>) 134900 tensor(4.8825, device='cuda:0', grad_fn=<NllLossBackward0>) 135000 tensor(4.7189, device='cuda:0', grad_fn=<NllLossBackward0>) 135100 tensor(4.6685, device='cuda:0', grad_fn=<NllLossBackward0>) 135200 tensor(4.7691, device='cuda:0', grad_fn=<NllLossBackward0>) 135300 tensor(4.7166, device='cuda:0', grad_fn=<NllLossBackward0>) 135400 tensor(4.6312, device='cuda:0', grad_fn=<NllLossBackward0>) 135500 tensor(4.9367, device='cuda:0', grad_fn=<NllLossBackward0>) 135600 tensor(4.8343, device='cuda:0', grad_fn=<NllLossBackward0>) 135700 tensor(4.8131, device='cuda:0', grad_fn=<NllLossBackward0>) 135800 tensor(4.7636, device='cuda:0', grad_fn=<NllLossBackward0>) 135900 tensor(4.6484, device='cuda:0', grad_fn=<NllLossBackward0>) 136000 tensor(4.7523, device='cuda:0', grad_fn=<NllLossBackward0>) 136100 tensor(4.8950, device='cuda:0', grad_fn=<NllLossBackward0>) 136200 tensor(4.9203, device='cuda:0', grad_fn=<NllLossBackward0>) 136300 tensor(4.8050, device='cuda:0', grad_fn=<NllLossBackward0>) 136400 tensor(4.8176, device='cuda:0', grad_fn=<NllLossBackward0>) 136500 tensor(4.4287, device='cuda:0', grad_fn=<NllLossBackward0>) 136600 tensor(4.5731, device='cuda:0', grad_fn=<NllLossBackward0>) 136700 tensor(4.7796, device='cuda:0', grad_fn=<NllLossBackward0>) 136800 tensor(5.1640, device='cuda:0', grad_fn=<NllLossBackward0>) 136900 tensor(4.5927, device='cuda:0', grad_fn=<NllLossBackward0>) 137000 tensor(4.5682, device='cuda:0', grad_fn=<NllLossBackward0>) 137100 tensor(5.0697, device='cuda:0', grad_fn=<NllLossBackward0>) 137200 tensor(4.8193, device='cuda:0', grad_fn=<NllLossBackward0>) 137300 tensor(4.8367, device='cuda:0', grad_fn=<NllLossBackward0>) 137400 tensor(4.9668, device='cuda:0', grad_fn=<NllLossBackward0>) 137500 tensor(4.6632, device='cuda:0', grad_fn=<NllLossBackward0>) 137600 tensor(5.2432, device='cuda:0', grad_fn=<NllLossBackward0>) 137700 tensor(4.6309, device='cuda:0', grad_fn=<NllLossBackward0>) 137800 tensor(4.7814, device='cuda:0', grad_fn=<NllLossBackward0>) 137900 tensor(4.4679, device='cuda:0', grad_fn=<NllLossBackward0>) 138000 tensor(4.9990, device='cuda:0', grad_fn=<NllLossBackward0>) 138100 tensor(4.6476, device='cuda:0', grad_fn=<NllLossBackward0>) 138200 tensor(4.9484, device='cuda:0', grad_fn=<NllLossBackward0>) 138300 tensor(4.8118, device='cuda:0', grad_fn=<NllLossBackward0>) 138400 tensor(4.8842, device='cuda:0', grad_fn=<NllLossBackward0>) 138500 tensor(4.7052, device='cuda:0', grad_fn=<NllLossBackward0>) 138600 tensor(3.9411, device='cuda:0', grad_fn=<NllLossBackward0>) 138700 tensor(5.0011, device='cuda:0', grad_fn=<NllLossBackward0>) 138800 tensor(4.8827, device='cuda:0', grad_fn=<NllLossBackward0>) 138900 tensor(4.8508, device='cuda:0', grad_fn=<NllLossBackward0>) 139000 tensor(4.6841, device='cuda:0', grad_fn=<NllLossBackward0>) Creating outputs in dev-0 Creating outputs in test-A Creating outputs in dev-0 Creating outputs in test-A Creating outputs in dev-0 Creating outputs in test-A