260 KiB
260 KiB
connect to google drive (working on colab)
from google.colab import drive
drive.mount('/content/drive')Mounted at /content/drive
!mkdir moj7
%cd drive/content/drive
%cd MyDrive/content/drive/MyDrive
%cd moj7/content/drive/MyDrive/moj7
!pwd/content/drive/MyDrive/moj7
Preprocess
import retrain_file ='train/in.tsv.xz'
test_file = 'test-A/in.tsv.xz'
out_file = 'test-A/out.tsv'
def preprocess(line):
line = replace_endline(line)
line = get_rid_of_header(line)
return line
def get_rid_of_header(line):
line = line.split('\t')[6:]
return "".join(line)
def replace_endline(line):
line = re.sub("\\\\n|\\\\+", " ", line)
return linefrom itertools import islice
import regex as re
import sys
from torchtext.vocab import build_vocab_from_iterator
import lzma
import pickle
def get_words_from_line(line):
line = line.rstrip()
yield '<s>'
line = preprocess(line)
for t in line.split(' '):
yield t
yield '</s>'
def get_word_lines_from_file(file_name):
n = 0
with lzma.open(file_name, 'r') as fh:
for line in fh:
n+=1
if n%1000==0:
print(n)
yield get_words_from_line(line.decode('utf-8'))
#vocab_size = 20000
vocab_size = 20000
vocab = build_vocab_from_iterator(
get_word_lines_from_file(train_file),
max_tokens = vocab_size,
specials = ['<unk>'])
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 13000 14000 15000 16000 17000 18000 19000 20000 21000 22000 23000 24000 25000 26000 27000 28000 29000 30000 31000 32000 33000 34000 35000 36000 37000 38000 39000 40000 41000 42000 43000 44000 45000 46000 47000 48000 49000 50000 51000 52000 53000 54000 55000 56000 57000 58000 59000 60000 61000 62000 63000 64000 65000 66000 67000 68000 69000 70000 71000 72000 73000 74000 75000 76000 77000 78000 79000 80000 81000 82000 83000 84000 85000 86000 87000 88000 89000 90000 91000 92000 93000 94000 95000 96000 97000 98000 99000 100000 101000 102000 103000 104000 105000 106000 107000 108000 109000 110000 111000 112000 113000 114000 115000 116000 117000 118000 119000 120000 121000 122000 123000 124000 125000 126000 127000 128000 129000 130000 131000 132000 133000 134000 135000 136000 137000 138000 139000 140000 141000 142000 143000 144000 145000 146000 147000 148000 149000 150000 151000 152000 153000 154000 155000 156000 157000 158000 159000 160000 161000 162000 163000 164000 165000 166000 167000 168000 169000 170000 171000 172000 173000 174000 175000 176000 177000 178000 179000 180000 181000 182000 183000 184000 185000 186000 187000 188000 189000 190000 191000 192000 193000 194000 195000 196000 197000 198000 199000 200000 201000 202000 203000 204000 205000 206000 207000 208000 209000 210000 211000 212000 213000 214000 215000 216000 217000 218000 219000 220000 221000 222000 223000 224000 225000 226000 227000 228000 229000 230000 231000 232000 233000 234000 235000 236000 237000 238000 239000 240000 241000 242000 243000 244000 245000 246000 247000 248000 249000 250000 251000 252000 253000 254000 255000 256000 257000 258000 259000 260000 261000 262000 263000 264000 265000 266000 267000 268000 269000 270000 271000 272000 273000 274000 275000 276000 277000 278000 279000 280000 281000 282000 283000 284000 285000 286000 287000 288000 289000 290000 291000 292000 293000 294000 295000 296000 297000 298000 299000 300000 301000 302000 303000 304000 305000 306000 307000 308000 309000 310000 311000 312000 313000 314000 315000 316000 317000 318000 319000 320000 321000 322000 323000 324000 325000 326000 327000 328000 329000 330000 331000 332000 333000 334000 335000 336000 337000 338000 339000 340000 341000 342000 343000 344000 345000 346000 347000 348000 349000 350000 351000 352000 353000 354000 355000 356000 357000 358000 359000 360000 361000 362000 363000 364000 365000 366000 367000 368000 369000 370000 371000 372000 373000 374000 375000 376000 377000 378000 379000 380000 381000 382000 383000 384000 385000 386000 387000 388000 389000 390000 391000 392000 393000 394000 395000 396000 397000 398000 399000 400000 401000 402000 403000 404000 405000 406000 407000 408000 409000 410000 411000 412000 413000 414000 415000 416000 417000 418000 419000 420000 421000 422000 423000 424000 425000 426000 427000 428000 429000 430000 431000 432000
vocab['no']50
with open('filename.pickle', 'wb') as handle:
pickle.dump(vocab, handle, protocol=pickle.HIGHEST_PROTOCOL)Create NN
from torch import nn
import torch
import pickle
# embed_size = 150
embed_size = 150
class Bigram(nn.Module):
def __init__(self, vocabulary_size, embedding_size):
super(Bigram, self).__init__()
self.model = nn.Sequential(
nn.Embedding(vocabulary_size, embedding_size),
nn.Linear(embedding_size, vocabulary_size),
nn.Softmax()
)
def forward(self, x):
return self.model(x)
model = Bigram(vocab_size, embed_size)
vocab.set_default_index(vocab['<unk>'])
res = torch.tensor(vocab.forward(['order']))
print(res)
tensor([215])
from torch.utils.data import IterableDataset
import itertools
def look_ahead_iterator(gen):
prev = None
for item in gen:
if prev is not None:
yield (prev, item)
prev = item
class Bigrams(IterableDataset):
def __init__(self, text_file, vocabulary_size):
self.vocab = build_vocab_from_iterator(
get_word_lines_from_file(text_file),
max_tokens = vocabulary_size,
specials = ['<unk>'])
self.vocab.set_default_index(self.vocab['<unk>'])
self.vocabulary_size = vocabulary_size
self.text_file = text_file
def __iter__(self):
return look_ahead_iterator(
(self.vocab[t] for t in itertools.chain.from_iterable(get_word_lines_from_file(self.text_file))))
train_dataset = Bigrams(train_file, vocab_size)1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000 13000 14000 15000 16000 17000 18000 19000 20000 21000 22000 23000 24000 25000 26000 27000 28000 29000 30000 31000 32000 33000 34000 35000 36000 37000 38000 39000 40000 41000 42000 43000 44000 45000 46000 47000 48000 49000 50000 51000 52000 53000 54000 55000 56000 57000 58000 59000 60000 61000 62000 63000 64000 65000 66000 67000 68000 69000 70000 71000 72000 73000 74000 75000 76000 77000 78000 79000 80000 81000 82000 83000 84000 85000 86000 87000 88000 89000 90000 91000 92000 93000 94000 95000 96000 97000 98000 99000 100000 101000 102000 103000 104000 105000 106000 107000 108000 109000 110000 111000 112000 113000 114000 115000 116000 117000 118000 119000 120000 121000 122000 123000 124000 125000 126000 127000 128000 129000 130000 131000 132000 133000 134000 135000 136000 137000 138000 139000 140000 141000 142000 143000 144000 145000 146000 147000 148000 149000 150000 151000 152000 153000 154000 155000 156000 157000 158000 159000 160000 161000 162000 163000 164000 165000 166000 167000 168000 169000 170000 171000 172000 173000 174000 175000 176000 177000 178000 179000 180000 181000 182000 183000 184000 185000 186000 187000 188000 189000 190000 191000 192000 193000 194000 195000 196000 197000 198000 199000 200000 201000 202000 203000 204000 205000 206000 207000 208000 209000 210000 211000 212000 213000 214000 215000 216000 217000 218000 219000 220000 221000 222000 223000 224000 225000 226000 227000 228000 229000 230000 231000 232000 233000 234000 235000 236000 237000 238000 239000 240000 241000 242000 243000 244000 245000 246000 247000 248000 249000 250000 251000 252000 253000 254000 255000 256000 257000 258000 259000 260000 261000 262000 263000 264000 265000 266000 267000 268000 269000 270000 271000 272000 273000 274000 275000 276000 277000 278000 279000 280000 281000 282000 283000 284000 285000 286000 287000 288000 289000 290000 291000 292000 293000 294000 295000 296000 297000 298000 299000 300000 301000 302000 303000 304000 305000 306000 307000 308000 309000 310000 311000 312000 313000 314000 315000 316000 317000 318000 319000 320000 321000 322000 323000 324000 325000 326000 327000 328000 329000 330000 331000 332000 333000 334000 335000 336000 337000 338000 339000 340000 341000 342000 343000 344000 345000 346000 347000 348000 349000 350000 351000 352000 353000 354000 355000 356000 357000 358000 359000 360000 361000 362000 363000 364000 365000 366000 367000 368000 369000 370000 371000 372000 373000 374000 375000 376000 377000 378000 379000 380000 381000 382000 383000 384000 385000 386000 387000 388000 389000 390000 391000 392000 393000 394000 395000 396000 397000 398000 399000 400000 401000 402000 403000 404000 405000 406000 407000 408000 409000 410000 411000 412000 413000 414000 415000 416000 417000 418000 419000 420000 421000 422000 423000 424000 425000 426000 427000 428000 429000 430000 431000 432000
from torch.utils.data import DataLoader
next(iter(DataLoader(train_dataset, batch_size=5)))[tensor([ 23, 191, 5791, 1, 112]), tensor([ 191, 5791, 1, 112, 159])]
Train
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = Bigram(vocab_size, embed_size).to(device)
data = DataLoader(train_dataset, batch_size=1000)
optimizer = torch.optim.Adam(model.parameters())
criterion = torch.nn.NLLLoss()
## epochs=2
for i in range(2):
print('epoch: =', i)
model.train()
step = 0
for x, y in data:
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()
optimizer.step()
torch.save(model.state_dict(), 'model.bin')
epoch: = 0
/usr/local/lib/python3.10/dist-packages/torch/nn/modules/container.py:217: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument. input = module(input)
0 tensor(10.3037, device='cuda:0', grad_fn=<NllLossBackward0>) 100 tensor(8.7506, device='cuda:0', grad_fn=<NllLossBackward0>) 200 tensor(7.8141, device='cuda:0', grad_fn=<NllLossBackward0>) 1000 300 tensor(7.4218, device='cuda:0', grad_fn=<NllLossBackward0>) 400 tensor(7.1627, device='cuda:0', grad_fn=<NllLossBackward0>) 500 tensor(6.7964, device='cuda:0', grad_fn=<NllLossBackward0>) 2000 600 tensor(6.4704, device='cuda:0', grad_fn=<NllLossBackward0>) 700 tensor(6.3798, device='cuda:0', grad_fn=<NllLossBackward0>) 800 tensor(6.2849, device='cuda:0', grad_fn=<NllLossBackward0>) 3000 900 tensor(6.3975, device='cuda:0', grad_fn=<NllLossBackward0>) 1000 tensor(6.0096, device='cuda:0', grad_fn=<NllLossBackward0>) 1100 tensor(5.7434, device='cuda:0', grad_fn=<NllLossBackward0>) 4000 1200 tensor(5.9602, device='cuda:0', grad_fn=<NllLossBackward0>) 1300 tensor(6.1623, device='cuda:0', grad_fn=<NllLossBackward0>) 1400 tensor(6.1647, device='cuda:0', grad_fn=<NllLossBackward0>) 5000 1500 tensor(6.1010, device='cuda:0', grad_fn=<NllLossBackward0>) 1600 tensor(6.0634, device='cuda:0', grad_fn=<NllLossBackward0>) 6000 1700 tensor(5.9149, device='cuda:0', grad_fn=<NllLossBackward0>) 1800 tensor(5.7918, device='cuda:0', grad_fn=<NllLossBackward0>) 1900 tensor(5.6739, device='cuda:0', grad_fn=<NllLossBackward0>) 7000 2000 tensor(5.5298, device='cuda:0', grad_fn=<NllLossBackward0>) 2100 tensor(5.8011, device='cuda:0', grad_fn=<NllLossBackward0>) 2200 tensor(5.4338, device='cuda:0', grad_fn=<NllLossBackward0>) 8000 2300 tensor(5.7522, device='cuda:0', grad_fn=<NllLossBackward0>) 2400 tensor(5.0313, device='cuda:0', grad_fn=<NllLossBackward0>) 2500 tensor(5.7116, device='cuda:0', grad_fn=<NllLossBackward0>) 9000 2600 tensor(5.2706, device='cuda:0', grad_fn=<NllLossBackward0>) 2700 tensor(5.6324, device='cuda:0', grad_fn=<NllLossBackward0>) 2800 tensor(5.0710, device='cuda:0', grad_fn=<NllLossBackward0>) 10000 2900 tensor(5.5921, device='cuda:0', grad_fn=<NllLossBackward0>) 3000 tensor(5.4808, device='cuda:0', grad_fn=<NllLossBackward0>) 11000 3100 tensor(5.3611, device='cuda:0', grad_fn=<NllLossBackward0>) 3200 tensor(5.6228, device='cuda:0', grad_fn=<NllLossBackward0>) 3300 tensor(5.4286, device='cuda:0', grad_fn=<NllLossBackward0>) 12000 3400 tensor(5.3550, device='cuda:0', grad_fn=<NllLossBackward0>) 3500 tensor(5.4032, device='cuda:0', grad_fn=<NllLossBackward0>) 3600 tensor(5.1070, device='cuda:0', grad_fn=<NllLossBackward0>) 13000 3700 tensor(5.4506, device='cuda:0', grad_fn=<NllLossBackward0>) 3800 tensor(5.4622, device='cuda:0', grad_fn=<NllLossBackward0>) 3900 tensor(5.4984, device='cuda:0', grad_fn=<NllLossBackward0>) 14000 4000 tensor(5.1740, device='cuda:0', grad_fn=<NllLossBackward0>) 4100 tensor(5.6064, device='cuda:0', grad_fn=<NllLossBackward0>) 4200 tensor(5.0705, device='cuda:0', grad_fn=<NllLossBackward0>) 15000 4300 tensor(5.5181, device='cuda:0', grad_fn=<NllLossBackward0>) 4400 tensor(5.2919, device='cuda:0', grad_fn=<NllLossBackward0>) 16000 4500 tensor(5.5021, device='cuda:0', grad_fn=<NllLossBackward0>) 4600 tensor(5.5308, device='cuda:0', grad_fn=<NllLossBackward0>) 4700 tensor(5.4699, device='cuda:0', grad_fn=<NllLossBackward0>) 17000 4800 tensor(5.2686, device='cuda:0', grad_fn=<NllLossBackward0>) 4900 tensor(5.4776, device='cuda:0', grad_fn=<NllLossBackward0>) 5000 tensor(5.5061, device='cuda:0', grad_fn=<NllLossBackward0>) 18000 5100 tensor(5.3180, device='cuda:0', grad_fn=<NllLossBackward0>) 5200 tensor(5.5524, device='cuda:0', grad_fn=<NllLossBackward0>) 5300 tensor(5.3481, device='cuda:0', grad_fn=<NllLossBackward0>) 19000 5400 tensor(5.2153, device='cuda:0', grad_fn=<NllLossBackward0>) 5500 tensor(5.4478, device='cuda:0', grad_fn=<NllLossBackward0>) 20000 5600 tensor(5.3441, device='cuda:0', grad_fn=<NllLossBackward0>) 5700 tensor(5.3958, device='cuda:0', grad_fn=<NllLossBackward0>) 5800 tensor(5.8945, device='cuda:0', grad_fn=<NllLossBackward0>) 21000 5900 tensor(5.5684, device='cuda:0', grad_fn=<NllLossBackward0>) 6000 tensor(5.5715, device='cuda:0', grad_fn=<NllLossBackward0>) 6100 tensor(5.2367, device='cuda:0', grad_fn=<NllLossBackward0>) 22000 6200 tensor(5.6976, device='cuda:0', grad_fn=<NllLossBackward0>) 6300 tensor(5.5367, device='cuda:0', grad_fn=<NllLossBackward0>) 6400 tensor(5.3024, device='cuda:0', grad_fn=<NllLossBackward0>) 23000 6500 tensor(5.3010, device='cuda:0', grad_fn=<NllLossBackward0>) 6600 tensor(6.0962, device='cuda:0', grad_fn=<NllLossBackward0>) 6700 tensor(5.0961, device='cuda:0', grad_fn=<NllLossBackward0>) 24000 6800 tensor(5.1091, device='cuda:0', grad_fn=<NllLossBackward0>) 6900 tensor(5.4123, device='cuda:0', grad_fn=<NllLossBackward0>) 25000 7000 tensor(5.3128, device='cuda:0', grad_fn=<NllLossBackward0>) 7100 tensor(5.3416, device='cuda:0', grad_fn=<NllLossBackward0>) 7200 tensor(5.4973, device='cuda:0', grad_fn=<NllLossBackward0>) 26000 7300 tensor(5.4418, device='cuda:0', grad_fn=<NllLossBackward0>) 7400 tensor(5.2171, device='cuda:0', grad_fn=<NllLossBackward0>) 7500 tensor(5.6509, device='cuda:0', grad_fn=<NllLossBackward0>) 27000 7600 tensor(5.0550, device='cuda:0', grad_fn=<NllLossBackward0>) 7700 tensor(5.4937, device='cuda:0', grad_fn=<NllLossBackward0>) 7800 tensor(5.9218, device='cuda:0', grad_fn=<NllLossBackward0>) 28000 7900 tensor(5.2853, device='cuda:0', grad_fn=<NllLossBackward0>) 8000 tensor(5.3146, device='cuda:0', grad_fn=<NllLossBackward0>) 8100 tensor(4.8552, device='cuda:0', grad_fn=<NllLossBackward0>) 29000 8200 tensor(5.3389, device='cuda:0', grad_fn=<NllLossBackward0>) 8300 tensor(5.2421, device='cuda:0', grad_fn=<NllLossBackward0>) 30000 8400 tensor(5.2460, device='cuda:0', grad_fn=<NllLossBackward0>) 8500 tensor(5.0331, device='cuda:0', grad_fn=<NllLossBackward0>) 8600 tensor(5.0050, device='cuda:0', grad_fn=<NllLossBackward0>) 31000 8700 tensor(5.3844, device='cuda:0', grad_fn=<NllLossBackward0>) 8800 tensor(5.4491, device='cuda:0', grad_fn=<NllLossBackward0>) 8900 tensor(5.6790, device='cuda:0', grad_fn=<NllLossBackward0>) 32000 9000 tensor(5.1118, device='cuda:0', grad_fn=<NllLossBackward0>) 9100 tensor(5.3567, device='cuda:0', grad_fn=<NllLossBackward0>) 9200 tensor(5.4141, device='cuda:0', grad_fn=<NllLossBackward0>) 33000 9300 tensor(5.3085, device='cuda:0', grad_fn=<NllLossBackward0>) 9400 tensor(5.2808, device='cuda:0', grad_fn=<NllLossBackward0>) 34000 9500 tensor(5.0931, device='cuda:0', grad_fn=<NllLossBackward0>) 9600 tensor(5.1090, device='cuda:0', grad_fn=<NllLossBackward0>) 9700 tensor(5.2519, device='cuda:0', grad_fn=<NllLossBackward0>) 35000 9800 tensor(5.3852, device='cuda:0', grad_fn=<NllLossBackward0>) 9900 tensor(5.0943, device='cuda:0', grad_fn=<NllLossBackward0>) 10000 tensor(5.4690, device='cuda:0', grad_fn=<NllLossBackward0>) 36000 10100 tensor(5.4348, device='cuda:0', grad_fn=<NllLossBackward0>) 10200 tensor(5.3262, device='cuda:0', grad_fn=<NllLossBackward0>) 10300 tensor(5.4878, device='cuda:0', grad_fn=<NllLossBackward0>) 37000 10400 tensor(5.2384, device='cuda:0', grad_fn=<NllLossBackward0>) 10500 tensor(5.2151, device='cuda:0', grad_fn=<NllLossBackward0>) 10600 tensor(4.8722, device='cuda:0', grad_fn=<NllLossBackward0>) 38000 10700 tensor(5.4325, device='cuda:0', grad_fn=<NllLossBackward0>) 10800 tensor(4.8699, device='cuda:0', grad_fn=<NllLossBackward0>) 39000 10900 tensor(5.3448, device='cuda:0', grad_fn=<NllLossBackward0>) 11000 tensor(5.1358, device='cuda:0', grad_fn=<NllLossBackward0>) 11100 tensor(5.0432, device='cuda:0', grad_fn=<NllLossBackward0>) 40000 11200 tensor(5.4062, device='cuda:0', grad_fn=<NllLossBackward0>) 11300 tensor(5.4040, device='cuda:0', grad_fn=<NllLossBackward0>) 11400 tensor(5.5312, device='cuda:0', grad_fn=<NllLossBackward0>) 41000 11500 tensor(5.4374, device='cuda:0', grad_fn=<NllLossBackward0>) 11600 tensor(5.0998, device='cuda:0', grad_fn=<NllLossBackward0>) 11700 tensor(5.4217, device='cuda:0', grad_fn=<NllLossBackward0>) 42000 11800 tensor(5.5747, device='cuda:0', grad_fn=<NllLossBackward0>) 11900 tensor(5.0467, device='cuda:0', grad_fn=<NllLossBackward0>) 12000 tensor(5.4270, device='cuda:0', grad_fn=<NllLossBackward0>) 43000 12100 tensor(5.2043, device='cuda:0', grad_fn=<NllLossBackward0>) 12200 tensor(5.2369, device='cuda:0', grad_fn=<NllLossBackward0>) 44000 12300 tensor(5.4465, device='cuda:0', grad_fn=<NllLossBackward0>) 12400 tensor(4.9839, device='cuda:0', grad_fn=<NllLossBackward0>) 12500 tensor(5.3214, device='cuda:0', grad_fn=<NllLossBackward0>) 45000 12600 tensor(5.1928, device='cuda:0', grad_fn=<NllLossBackward0>) 12700 tensor(4.9646, device='cuda:0', grad_fn=<NllLossBackward0>) 12800 tensor(5.3325, device='cuda:0', grad_fn=<NllLossBackward0>) 46000 12900 tensor(5.4429, device='cuda:0', grad_fn=<NllLossBackward0>) 13000 tensor(5.0652, device='cuda:0', grad_fn=<NllLossBackward0>) 13100 tensor(5.3126, device='cuda:0', grad_fn=<NllLossBackward0>) 47000 13200 tensor(5.4124, device='cuda:0', grad_fn=<NllLossBackward0>) 13300 tensor(5.5385, device='cuda:0', grad_fn=<NllLossBackward0>) 13400 tensor(5.0986, device='cuda:0', grad_fn=<NllLossBackward0>) 48000 13500 tensor(5.2693, device='cuda:0', grad_fn=<NllLossBackward0>) 13600 tensor(5.2136, device='cuda:0', grad_fn=<NllLossBackward0>) 49000 13700 tensor(5.5169, device='cuda:0', grad_fn=<NllLossBackward0>) 13800 tensor(5.1840, device='cuda:0', grad_fn=<NllLossBackward0>) 13900 tensor(5.2700, device='cuda:0', grad_fn=<NllLossBackward0>) 50000 14000 tensor(5.2077, device='cuda:0', grad_fn=<NllLossBackward0>) 14100 tensor(5.3791, device='cuda:0', grad_fn=<NllLossBackward0>) 14200 tensor(5.4008, device='cuda:0', grad_fn=<NllLossBackward0>) 51000 14300 tensor(5.3506, device='cuda:0', grad_fn=<NllLossBackward0>) 14400 tensor(4.7662, device='cuda:0', grad_fn=<NllLossBackward0>) 14500 tensor(4.9474, device='cuda:0', grad_fn=<NllLossBackward0>) 52000 14600 tensor(5.0245, device='cuda:0', grad_fn=<NllLossBackward0>) 14700 tensor(5.3977, device='cuda:0', grad_fn=<NllLossBackward0>) 14800 tensor(4.9653, device='cuda:0', grad_fn=<NllLossBackward0>) 53000 14900 tensor(4.8947, device='cuda:0', grad_fn=<NllLossBackward0>) 15000 tensor(5.3548, device='cuda:0', grad_fn=<NllLossBackward0>) 54000 15100 tensor(4.7244, device='cuda:0', grad_fn=<NllLossBackward0>) 15200 tensor(4.9752, device='cuda:0', grad_fn=<NllLossBackward0>) 15300 tensor(5.3929, device='cuda:0', grad_fn=<NllLossBackward0>) 55000 15400 tensor(5.3096, device='cuda:0', grad_fn=<NllLossBackward0>) 15500 tensor(5.1247, device='cuda:0', grad_fn=<NllLossBackward0>) 15600 tensor(5.2753, device='cuda:0', grad_fn=<NllLossBackward0>) 56000 15700 tensor(5.2373, device='cuda:0', grad_fn=<NllLossBackward0>) 15800 tensor(4.9997, device='cuda:0', grad_fn=<NllLossBackward0>) 15900 tensor(5.1718, device='cuda:0', grad_fn=<NllLossBackward0>) 57000 16000 tensor(5.5952, device='cuda:0', grad_fn=<NllLossBackward0>) 16100 tensor(5.3699, device='cuda:0', grad_fn=<NllLossBackward0>) 16200 tensor(5.0923, device='cuda:0', grad_fn=<NllLossBackward0>) 58000 16300 tensor(4.9985, device='cuda:0', grad_fn=<NllLossBackward0>) 16400 tensor(5.3076, device='cuda:0', grad_fn=<NllLossBackward0>) 59000 16500 tensor(5.1994, device='cuda:0', grad_fn=<NllLossBackward0>) 16600 tensor(5.3672, device='cuda:0', grad_fn=<NllLossBackward0>) 16700 tensor(5.2054, device='cuda:0', grad_fn=<NllLossBackward0>) 60000 16800 tensor(5.3379, device='cuda:0', grad_fn=<NllLossBackward0>) 16900 tensor(5.2785, device='cuda:0', grad_fn=<NllLossBackward0>) 17000 tensor(5.2590, device='cuda:0', grad_fn=<NllLossBackward0>) 61000 17100 tensor(5.3564, device='cuda:0', grad_fn=<NllLossBackward0>) 17200 tensor(5.3598, device='cuda:0', grad_fn=<NllLossBackward0>) 17300 tensor(4.7786, device='cuda:0', grad_fn=<NllLossBackward0>) 62000 17400 tensor(5.2639, device='cuda:0', grad_fn=<NllLossBackward0>) 17500 tensor(5.2037, device='cuda:0', grad_fn=<NllLossBackward0>) 17600 tensor(5.1158, device='cuda:0', grad_fn=<NllLossBackward0>) 63000 17700 tensor(4.9831, device='cuda:0', grad_fn=<NllLossBackward0>) 17800 tensor(4.8950, device='cuda:0', grad_fn=<NllLossBackward0>) 64000 17900 tensor(5.0928, device='cuda:0', grad_fn=<NllLossBackward0>) 18000 tensor(5.3423, device='cuda:0', grad_fn=<NllLossBackward0>) 18100 tensor(5.1760, device='cuda:0', grad_fn=<NllLossBackward0>) 65000 18200 tensor(5.2021, device='cuda:0', grad_fn=<NllLossBackward0>) 18300 tensor(5.1306, device='cuda:0', grad_fn=<NllLossBackward0>) 18400 tensor(5.1199, device='cuda:0', grad_fn=<NllLossBackward0>) 66000 18500 tensor(5.2082, device='cuda:0', grad_fn=<NllLossBackward0>) 18600 tensor(5.3290, device='cuda:0', grad_fn=<NllLossBackward0>) 18700 tensor(5.2257, device='cuda:0', grad_fn=<NllLossBackward0>) 67000 18800 tensor(4.9107, device='cuda:0', grad_fn=<NllLossBackward0>) 18900 tensor(5.3400, device='cuda:0', grad_fn=<NllLossBackward0>) 68000 19000 tensor(5.1366, device='cuda:0', grad_fn=<NllLossBackward0>) 19100 tensor(5.1199, device='cuda:0', grad_fn=<NllLossBackward0>) 19200 tensor(5.2202, device='cuda:0', grad_fn=<NllLossBackward0>) 69000 19300 tensor(5.2236, device='cuda:0', grad_fn=<NllLossBackward0>) 19400 tensor(5.2953, device='cuda:0', grad_fn=<NllLossBackward0>) 19500 tensor(5.1308, device='cuda:0', grad_fn=<NllLossBackward0>) 70000 19600 tensor(5.3578, device='cuda:0', grad_fn=<NllLossBackward0>) 19700 tensor(5.1600, device='cuda:0', grad_fn=<NllLossBackward0>) 19800 tensor(4.6220, device='cuda:0', grad_fn=<NllLossBackward0>) 71000 19900 tensor(5.3731, device='cuda:0', grad_fn=<NllLossBackward0>) 20000 tensor(4.9936, device='cuda:0', grad_fn=<NllLossBackward0>) 20100 tensor(5.0817, device='cuda:0', grad_fn=<NllLossBackward0>) 72000 20200 tensor(5.1613, device='cuda:0', grad_fn=<NllLossBackward0>) 20300 tensor(5.3877, device='cuda:0', grad_fn=<NllLossBackward0>) 73000 20400 tensor(5.4114, device='cuda:0', grad_fn=<NllLossBackward0>) 20500 tensor(5.2609, device='cuda:0', grad_fn=<NllLossBackward0>) 20600 tensor(5.1378, device='cuda:0', grad_fn=<NllLossBackward0>) 74000 20700 tensor(5.0799, device='cuda:0', grad_fn=<NllLossBackward0>) 20800 tensor(5.3615, device='cuda:0', grad_fn=<NllLossBackward0>) 20900 tensor(5.3365, device='cuda:0', grad_fn=<NllLossBackward0>) 75000 21000 tensor(4.9244, device='cuda:0', grad_fn=<NllLossBackward0>) 21100 tensor(5.5084, device='cuda:0', grad_fn=<NllLossBackward0>) 21200 tensor(4.8769, device='cuda:0', grad_fn=<NllLossBackward0>) 76000 21300 tensor(5.3414, device='cuda:0', grad_fn=<NllLossBackward0>) 21400 tensor(5.0648, device='cuda:0', grad_fn=<NllLossBackward0>) 21500 tensor(5.0594, device='cuda:0', grad_fn=<NllLossBackward0>) 77000 21600 tensor(5.2537, device='cuda:0', grad_fn=<NllLossBackward0>) 21700 tensor(5.1834, device='cuda:0', grad_fn=<NllLossBackward0>) 21800 tensor(4.8151, device='cuda:0', grad_fn=<NllLossBackward0>) 78000 21900 tensor(5.3335, device='cuda:0', grad_fn=<NllLossBackward0>) 22000 tensor(4.9580, device='cuda:0', grad_fn=<NllLossBackward0>) 79000 22100 tensor(5.2262, device='cuda:0', grad_fn=<NllLossBackward0>) 22200 tensor(5.1946, device='cuda:0', grad_fn=<NllLossBackward0>) 22300 tensor(5.2404, device='cuda:0', grad_fn=<NllLossBackward0>) 80000 22400 tensor(4.9491, device='cuda:0', grad_fn=<NllLossBackward0>) 22500 tensor(4.6901, device='cuda:0', grad_fn=<NllLossBackward0>) 22600 tensor(5.1937, device='cuda:0', grad_fn=<NllLossBackward0>) 81000 22700 tensor(4.9937, device='cuda:0', grad_fn=<NllLossBackward0>) 22800 tensor(5.1401, device='cuda:0', grad_fn=<NllLossBackward0>) 22900 tensor(5.0599, device='cuda:0', grad_fn=<NllLossBackward0>) 82000 23000 tensor(5.4315, device='cuda:0', grad_fn=<NllLossBackward0>) 23100 tensor(5.1854, device='cuda:0', grad_fn=<NllLossBackward0>) 83000 23200 tensor(5.1033, device='cuda:0', grad_fn=<NllLossBackward0>) 23300 tensor(5.2352, device='cuda:0', grad_fn=<NllLossBackward0>) 23400 tensor(5.2004, device='cuda:0', grad_fn=<NllLossBackward0>) 84000 23500 tensor(5.0866, device='cuda:0', grad_fn=<NllLossBackward0>) 23600 tensor(5.2372, device='cuda:0', grad_fn=<NllLossBackward0>) 23700 tensor(5.4711, device='cuda:0', grad_fn=<NllLossBackward0>) 85000 23800 tensor(5.4030, device='cuda:0', grad_fn=<NllLossBackward0>) 23900 tensor(5.3589, device='cuda:0', grad_fn=<NllLossBackward0>) 24000 tensor(5.1646, device='cuda:0', grad_fn=<NllLossBackward0>) 86000 24100 tensor(5.4865, device='cuda:0', grad_fn=<NllLossBackward0>) 24200 tensor(5.3663, device='cuda:0', grad_fn=<NllLossBackward0>) 24300 tensor(5.1760, device='cuda:0', grad_fn=<NllLossBackward0>) 87000 24400 tensor(5.2950, device='cuda:0', grad_fn=<NllLossBackward0>) 24500 tensor(5.0376, device='cuda:0', grad_fn=<NllLossBackward0>) 88000 24600 tensor(5.1229, device='cuda:0', grad_fn=<NllLossBackward0>) 24700 tensor(5.3261, device='cuda:0', grad_fn=<NllLossBackward0>) 24800 tensor(5.3953, device='cuda:0', grad_fn=<NllLossBackward0>) 89000 24900 tensor(5.2734, device='cuda:0', grad_fn=<NllLossBackward0>) 25000 tensor(5.5544, device='cuda:0', grad_fn=<NllLossBackward0>) 25100 tensor(5.1847, device='cuda:0', grad_fn=<NllLossBackward0>) 90000 25200 tensor(5.2963, device='cuda:0', grad_fn=<NllLossBackward0>) 25300 tensor(5.2350, device='cuda:0', grad_fn=<NllLossBackward0>) 25400 tensor(5.1483, device='cuda:0', grad_fn=<NllLossBackward0>) 91000 25500 tensor(5.0571, device='cuda:0', grad_fn=<NllLossBackward0>) 25600 tensor(4.8022, device='cuda:0', grad_fn=<NllLossBackward0>) 25700 tensor(5.2115, device='cuda:0', grad_fn=<NllLossBackward0>) 92000 25800 tensor(5.1675, device='cuda:0', grad_fn=<NllLossBackward0>) 25900 tensor(5.2786, device='cuda:0', grad_fn=<NllLossBackward0>) 93000 26000 tensor(4.8893, device='cuda:0', grad_fn=<NllLossBackward0>) 26100 tensor(5.3295, device='cuda:0', grad_fn=<NllLossBackward0>) 26200 tensor(4.9849, device='cuda:0', grad_fn=<NllLossBackward0>) 94000 26300 tensor(5.2289, device='cuda:0', grad_fn=<NllLossBackward0>) 26400 tensor(5.1430, device='cuda:0', grad_fn=<NllLossBackward0>) 26500 tensor(5.1389, device='cuda:0', grad_fn=<NllLossBackward0>) 95000 26600 tensor(5.3339, device='cuda:0', grad_fn=<NllLossBackward0>) 26700 tensor(5.1904, device='cuda:0', grad_fn=<NllLossBackward0>) 26800 tensor(5.0073, device='cuda:0', grad_fn=<NllLossBackward0>) 96000 26900 tensor(5.1798, device='cuda:0', grad_fn=<NllLossBackward0>) 27000 tensor(4.9300, device='cuda:0', grad_fn=<NllLossBackward0>) 27100 tensor(5.1331, device='cuda:0', grad_fn=<NllLossBackward0>) 97000 27200 tensor(5.1750, device='cuda:0', grad_fn=<NllLossBackward0>) 27300 tensor(5.1801, device='cuda:0', grad_fn=<NllLossBackward0>) 98000 27400 tensor(5.0430, device='cuda:0', grad_fn=<NllLossBackward0>) 27500 tensor(5.3107, device='cuda:0', grad_fn=<NllLossBackward0>) 27600 tensor(5.2876, device='cuda:0', grad_fn=<NllLossBackward0>) 99000 27700 tensor(5.2763, device='cuda:0', grad_fn=<NllLossBackward0>) 27800 tensor(5.0737, device='cuda:0', grad_fn=<NllLossBackward0>) 27900 tensor(5.0593, device='cuda:0', grad_fn=<NllLossBackward0>) 100000 28000 tensor(5.4758, device='cuda:0', grad_fn=<NllLossBackward0>) 28100 tensor(5.4440, device='cuda:0', grad_fn=<NllLossBackward0>) 28200 tensor(5.0862, device='cuda:0', grad_fn=<NllLossBackward0>) 101000 28300 tensor(5.2405, device='cuda:0', grad_fn=<NllLossBackward0>) 28400 tensor(4.8600, device='cuda:0', grad_fn=<NllLossBackward0>) 28500 tensor(5.2056, device='cuda:0', grad_fn=<NllLossBackward0>) 102000 28600 tensor(5.6196, device='cuda:0', grad_fn=<NllLossBackward0>) 28700 tensor(5.2235, device='cuda:0', grad_fn=<NllLossBackward0>) 103000 28800 tensor(5.2124, device='cuda:0', grad_fn=<NllLossBackward0>) 28900 tensor(4.8836, device='cuda:0', grad_fn=<NllLossBackward0>) 29000 tensor(4.9913, device='cuda:0', grad_fn=<NllLossBackward0>) 104000 29100 tensor(5.2689, device='cuda:0', grad_fn=<NllLossBackward0>) 29200 tensor(4.9352, device='cuda:0', grad_fn=<NllLossBackward0>) 29300 tensor(5.3143, device='cuda:0', grad_fn=<NllLossBackward0>) 105000 29400 tensor(5.2134, device='cuda:0', grad_fn=<NllLossBackward0>) 29500 tensor(5.0963, device='cuda:0', grad_fn=<NllLossBackward0>) 29600 tensor(5.4291, device='cuda:0', grad_fn=<NllLossBackward0>) 106000 29700 tensor(4.8093, device='cuda:0', grad_fn=<NllLossBackward0>) 29800 tensor(4.9271, device='cuda:0', grad_fn=<NllLossBackward0>) 29900 tensor(5.0479, device='cuda:0', grad_fn=<NllLossBackward0>) 107000 30000 tensor(5.2494, device='cuda:0', grad_fn=<NllLossBackward0>) 30100 tensor(5.2027, device='cuda:0', grad_fn=<NllLossBackward0>) 108000 30200 tensor(5.2714, device='cuda:0', grad_fn=<NllLossBackward0>) 30300 tensor(4.9223, device='cuda:0', grad_fn=<NllLossBackward0>) 30400 tensor(4.8816, device='cuda:0', grad_fn=<NllLossBackward0>) 109000 30500 tensor(4.9169, device='cuda:0', grad_fn=<NllLossBackward0>) 30600 tensor(5.1237, device='cuda:0', grad_fn=<NllLossBackward0>) 30700 tensor(5.1190, device='cuda:0', grad_fn=<NllLossBackward0>) 110000 30800 tensor(5.2271, device='cuda:0', grad_fn=<NllLossBackward0>) 30900 tensor(5.0194, device='cuda:0', grad_fn=<NllLossBackward0>) 31000 tensor(5.2287, device='cuda:0', grad_fn=<NllLossBackward0>) 111000 31100 tensor(5.3456, device='cuda:0', grad_fn=<NllLossBackward0>) 31200 tensor(5.0911, device='cuda:0', grad_fn=<NllLossBackward0>) 31300 tensor(5.3793, device='cuda:0', grad_fn=<NllLossBackward0>) 112000 31400 tensor(5.0410, device='cuda:0', grad_fn=<NllLossBackward0>) 31500 tensor(5.0014, device='cuda:0', grad_fn=<NllLossBackward0>) 113000 31600 tensor(4.9323, device='cuda:0', grad_fn=<NllLossBackward0>) 31700 tensor(5.3869, device='cuda:0', grad_fn=<NllLossBackward0>) 31800 tensor(4.8621, device='cuda:0', grad_fn=<NllLossBackward0>) 114000 31900 tensor(5.1184, device='cuda:0', grad_fn=<NllLossBackward0>) 32000 tensor(5.2873, device='cuda:0', grad_fn=<NllLossBackward0>) 32100 tensor(4.8723, device='cuda:0', grad_fn=<NllLossBackward0>) 115000 32200 tensor(5.4421, device='cuda:0', grad_fn=<NllLossBackward0>) 32300 tensor(5.0732, device='cuda:0', grad_fn=<NllLossBackward0>) 32400 tensor(5.0461, device='cuda:0', grad_fn=<NllLossBackward0>) 116000 32500 tensor(5.3476, device='cuda:0', grad_fn=<NllLossBackward0>) 32600 tensor(5.1149, device='cuda:0', grad_fn=<NllLossBackward0>) 117000 32700 tensor(5.2261, device='cuda:0', grad_fn=<NllLossBackward0>) 32800 tensor(5.1622, device='cuda:0', grad_fn=<NllLossBackward0>) 32900 tensor(5.0429, device='cuda:0', grad_fn=<NllLossBackward0>) 118000 33000 tensor(5.1329, device='cuda:0', grad_fn=<NllLossBackward0>) 33100 tensor(5.2284, device='cuda:0', grad_fn=<NllLossBackward0>) 33200 tensor(5.1794, device='cuda:0', grad_fn=<NllLossBackward0>) 119000 33300 tensor(4.8535, device='cuda:0', grad_fn=<NllLossBackward0>) 33400 tensor(4.7420, device='cuda:0', grad_fn=<NllLossBackward0>) 33500 tensor(5.4625, device='cuda:0', grad_fn=<NllLossBackward0>) 120000 33600 tensor(4.9975, device='cuda:0', grad_fn=<NllLossBackward0>) 33700 tensor(5.2172, device='cuda:0', grad_fn=<NllLossBackward0>) 33800 tensor(5.1135, device='cuda:0', grad_fn=<NllLossBackward0>) 121000 33900 tensor(5.1974, device='cuda:0', grad_fn=<NllLossBackward0>) 34000 tensor(5.0866, device='cuda:0', grad_fn=<NllLossBackward0>) 122000 34100 tensor(5.1738, device='cuda:0', grad_fn=<NllLossBackward0>) 34200 tensor(5.2338, device='cuda:0', grad_fn=<NllLossBackward0>) 34300 tensor(5.3866, device='cuda:0', grad_fn=<NllLossBackward0>) 123000 34400 tensor(5.2937, device='cuda:0', grad_fn=<NllLossBackward0>) 34500 tensor(5.4265, device='cuda:0', grad_fn=<NllLossBackward0>) 34600 tensor(5.0077, device='cuda:0', grad_fn=<NllLossBackward0>) 124000 34700 tensor(5.3335, device='cuda:0', grad_fn=<NllLossBackward0>) 34800 tensor(4.9879, device='cuda:0', grad_fn=<NllLossBackward0>) 34900 tensor(5.3270, device='cuda:0', grad_fn=<NllLossBackward0>) 125000 35000 tensor(5.2056, device='cuda:0', grad_fn=<NllLossBackward0>) 35100 tensor(5.1341, device='cuda:0', grad_fn=<NllLossBackward0>) 35200 tensor(5.2629, device='cuda:0', grad_fn=<NllLossBackward0>) 126000 35300 tensor(5.0193, device='cuda:0', grad_fn=<NllLossBackward0>) 35400 tensor(5.2414, device='cuda:0', grad_fn=<NllLossBackward0>) 127000 35500 tensor(5.3116, device='cuda:0', grad_fn=<NllLossBackward0>) 35600 tensor(5.3802, device='cuda:0', grad_fn=<NllLossBackward0>) 35700 tensor(5.5126, device='cuda:0', grad_fn=<NllLossBackward0>) 128000 35800 tensor(5.2014, device='cuda:0', grad_fn=<NllLossBackward0>) 35900 tensor(5.2182, device='cuda:0', grad_fn=<NllLossBackward0>) 36000 tensor(5.2556, device='cuda:0', grad_fn=<NllLossBackward0>) 129000 36100 tensor(5.2955, device='cuda:0', grad_fn=<NllLossBackward0>) 36200 tensor(5.2068, device='cuda:0', grad_fn=<NllLossBackward0>) 36300 tensor(5.4835, device='cuda:0', grad_fn=<NllLossBackward0>) 130000 36400 tensor(5.2844, device='cuda:0', grad_fn=<NllLossBackward0>) 36500 tensor(5.4568, device='cuda:0', grad_fn=<NllLossBackward0>) 131000 36600 tensor(4.9546, device='cuda:0', grad_fn=<NllLossBackward0>) 36700 tensor(5.5436, device='cuda:0', grad_fn=<NllLossBackward0>) 36800 tensor(5.4078, device='cuda:0', grad_fn=<NllLossBackward0>) 132000 36900 tensor(4.9905, device='cuda:0', grad_fn=<NllLossBackward0>) 37000 tensor(5.2532, device='cuda:0', grad_fn=<NllLossBackward0>) 37100 tensor(4.9246, device='cuda:0', grad_fn=<NllLossBackward0>) 133000 37200 tensor(4.8595, device='cuda:0', grad_fn=<NllLossBackward0>) 37300 tensor(4.9436, device='cuda:0', grad_fn=<NllLossBackward0>) 37400 tensor(4.9957, device='cuda:0', grad_fn=<NllLossBackward0>) 134000 37500 tensor(4.8686, device='cuda:0', grad_fn=<NllLossBackward0>) 37600 tensor(4.8965, device='cuda:0', grad_fn=<NllLossBackward0>) 37700 tensor(5.1955, device='cuda:0', grad_fn=<NllLossBackward0>) 135000 37800 tensor(5.2660, device='cuda:0', grad_fn=<NllLossBackward0>) 37900 tensor(5.3396, device='cuda:0', grad_fn=<NllLossBackward0>) 136000 38000 tensor(5.0582, device='cuda:0', grad_fn=<NllLossBackward0>) 38100 tensor(5.2351, device='cuda:0', grad_fn=<NllLossBackward0>) 38200 tensor(4.9177, device='cuda:0', grad_fn=<NllLossBackward0>) 137000 38300 tensor(5.3007, device='cuda:0', grad_fn=<NllLossBackward0>) 38400 tensor(4.9814, device='cuda:0', grad_fn=<NllLossBackward0>) 38500 tensor(5.2672, device='cuda:0', grad_fn=<NllLossBackward0>) 138000 38600 tensor(4.8161, device='cuda:0', grad_fn=<NllLossBackward0>) 38700 tensor(5.1975, device='cuda:0', grad_fn=<NllLossBackward0>) 38800 tensor(4.9423, device='cuda:0', grad_fn=<NllLossBackward0>) 139000 38900 tensor(5.0342, device='cuda:0', grad_fn=<NllLossBackward0>) 39000 tensor(5.2703, device='cuda:0', grad_fn=<NllLossBackward0>) 39100 tensor(4.9539, device='cuda:0', grad_fn=<NllLossBackward0>) 140000 39200 tensor(5.4193, device='cuda:0', grad_fn=<NllLossBackward0>) 39300 tensor(5.1109, device='cuda:0', grad_fn=<NllLossBackward0>) 141000 39400 tensor(5.0914, device='cuda:0', grad_fn=<NllLossBackward0>) 39500 tensor(5.3126, device='cuda:0', grad_fn=<NllLossBackward0>) 39600 tensor(4.9927, device='cuda:0', grad_fn=<NllLossBackward0>) 142000 39700 tensor(5.3725, device='cuda:0', grad_fn=<NllLossBackward0>) 39800 tensor(5.2926, device='cuda:0', grad_fn=<NllLossBackward0>) 39900 tensor(4.8999, device='cuda:0', grad_fn=<NllLossBackward0>) 143000 40000 tensor(5.0333, device='cuda:0', grad_fn=<NllLossBackward0>) 40100 tensor(5.1059, device='cuda:0', grad_fn=<NllLossBackward0>) 40200 tensor(5.1685, device='cuda:0', grad_fn=<NllLossBackward0>) 144000 40300 tensor(5.1375, device='cuda:0', grad_fn=<NllLossBackward0>) 40400 tensor(5.3952, device='cuda:0', grad_fn=<NllLossBackward0>) 40500 tensor(4.8467, device='cuda:0', grad_fn=<NllLossBackward0>) 145000 40600 tensor(5.4346, device='cuda:0', grad_fn=<NllLossBackward0>) 40700 tensor(5.2852, device='cuda:0', grad_fn=<NllLossBackward0>) 146000 40800 tensor(5.0129, device='cuda:0', grad_fn=<NllLossBackward0>) 40900 tensor(5.0657, device='cuda:0', grad_fn=<NllLossBackward0>) 41000 tensor(5.1874, device='cuda:0', grad_fn=<NllLossBackward0>) 147000 41100 tensor(5.0094, device='cuda:0', grad_fn=<NllLossBackward0>) 41200 tensor(5.2361, device='cuda:0', grad_fn=<NllLossBackward0>) 41300 tensor(4.8607, device='cuda:0', grad_fn=<NllLossBackward0>) 148000 41400 tensor(4.9584, device='cuda:0', grad_fn=<NllLossBackward0>) 41500 tensor(5.3754, device='cuda:0', grad_fn=<NllLossBackward0>) 41600 tensor(4.9372, device='cuda:0', grad_fn=<NllLossBackward0>) 149000 41700 tensor(5.2398, device='cuda:0', grad_fn=<NllLossBackward0>) 41800 tensor(5.1772, device='cuda:0', grad_fn=<NllLossBackward0>) 41900 tensor(5.1525, device='cuda:0', grad_fn=<NllLossBackward0>) 150000 42000 tensor(5.2827, device='cuda:0', grad_fn=<NllLossBackward0>) 42100 tensor(4.6855, device='cuda:0', grad_fn=<NllLossBackward0>) 151000 42200 tensor(5.0215, device='cuda:0', grad_fn=<NllLossBackward0>) 42300 tensor(5.1695, device='cuda:0', grad_fn=<NllLossBackward0>) 42400 tensor(5.0091, device='cuda:0', grad_fn=<NllLossBackward0>) 152000 42500 tensor(5.1274, device='cuda:0', grad_fn=<NllLossBackward0>) 42600 tensor(4.6771, device='cuda:0', grad_fn=<NllLossBackward0>) 42700 tensor(5.3160, device='cuda:0', grad_fn=<NllLossBackward0>) 153000 42800 tensor(5.2484, device='cuda:0', grad_fn=<NllLossBackward0>) 42900 tensor(5.0599, device='cuda:0', grad_fn=<NllLossBackward0>) 43000 tensor(5.2023, device='cuda:0', grad_fn=<NllLossBackward0>) 154000 43100 tensor(4.9059, device='cuda:0', grad_fn=<NllLossBackward0>) 43200 tensor(5.0727, device='cuda:0', grad_fn=<NllLossBackward0>) 43300 tensor(5.1986, device='cuda:0', grad_fn=<NllLossBackward0>) 155000 43400 tensor(5.1041, device='cuda:0', grad_fn=<NllLossBackward0>) 43500 tensor(5.4856, device='cuda:0', grad_fn=<NllLossBackward0>) 156000 43600 tensor(5.3947, device='cuda:0', grad_fn=<NllLossBackward0>) 43700 tensor(5.0493, device='cuda:0', grad_fn=<NllLossBackward0>) 43800 tensor(5.1703, device='cuda:0', grad_fn=<NllLossBackward0>) 157000 43900 tensor(4.9899, device='cuda:0', grad_fn=<NllLossBackward0>) 44000 tensor(4.8695, device='cuda:0', grad_fn=<NllLossBackward0>) 44100 tensor(5.3520, device='cuda:0', grad_fn=<NllLossBackward0>) 158000 44200 tensor(4.9379, device='cuda:0', grad_fn=<NllLossBackward0>) 44300 tensor(5.4677, device='cuda:0', grad_fn=<NllLossBackward0>) 44400 tensor(5.0739, device='cuda:0', grad_fn=<NllLossBackward0>) 159000 44500 tensor(4.9127, device='cuda:0', grad_fn=<NllLossBackward0>) 44600 tensor(5.0247, device='cuda:0', grad_fn=<NllLossBackward0>) 160000 44700 tensor(4.9641, device='cuda:0', grad_fn=<NllLossBackward0>) 44800 tensor(5.2538, device='cuda:0', grad_fn=<NllLossBackward0>) 44900 tensor(5.0657, device='cuda:0', grad_fn=<NllLossBackward0>) 161000 45000 tensor(4.9174, device='cuda:0', grad_fn=<NllLossBackward0>) 45100 tensor(5.2311, device='cuda:0', grad_fn=<NllLossBackward0>) 45200 tensor(5.0617, device='cuda:0', grad_fn=<NllLossBackward0>) 162000 45300 tensor(5.2111, device='cuda:0', grad_fn=<NllLossBackward0>) 45400 tensor(4.6137, device='cuda:0', grad_fn=<NllLossBackward0>) 45500 tensor(4.9275, device='cuda:0', grad_fn=<NllLossBackward0>) 163000 45600 tensor(5.2501, device='cuda:0', grad_fn=<NllLossBackward0>) 45700 tensor(5.2526, device='cuda:0', grad_fn=<NllLossBackward0>) 45800 tensor(5.1342, device='cuda:0', grad_fn=<NllLossBackward0>) 164000 45900 tensor(4.9400, device='cuda:0', grad_fn=<NllLossBackward0>) 46000 tensor(5.1521, device='cuda:0', grad_fn=<NllLossBackward0>) 165000 46100 tensor(5.0033, device='cuda:0', grad_fn=<NllLossBackward0>) 46200 tensor(4.9825, device='cuda:0', grad_fn=<NllLossBackward0>) 46300 tensor(5.3210, device='cuda:0', grad_fn=<NllLossBackward0>) 166000 46400 tensor(5.3635, device='cuda:0', grad_fn=<NllLossBackward0>) 46500 tensor(4.9169, device='cuda:0', grad_fn=<NllLossBackward0>) 46600 tensor(4.9497, device='cuda:0', grad_fn=<NllLossBackward0>) 167000 46700 tensor(5.2150, device='cuda:0', grad_fn=<NllLossBackward0>) 46800 tensor(4.8570, device='cuda:0', grad_fn=<NllLossBackward0>) 46900 tensor(4.9026, device='cuda:0', grad_fn=<NllLossBackward0>) 168000 47000 tensor(5.1486, device='cuda:0', grad_fn=<NllLossBackward0>) 47100 tensor(5.4026, device='cuda:0', grad_fn=<NllLossBackward0>) 47200 tensor(5.1028, device='cuda:0', grad_fn=<NllLossBackward0>) 169000 47300 tensor(5.2411, device='cuda:0', grad_fn=<NllLossBackward0>) 47400 tensor(4.9942, device='cuda:0', grad_fn=<NllLossBackward0>) 170000 47500 tensor(5.1984, device='cuda:0', grad_fn=<NllLossBackward0>) 47600 tensor(5.1100, device='cuda:0', grad_fn=<NllLossBackward0>) 47700 tensor(5.1202, device='cuda:0', grad_fn=<NllLossBackward0>) 171000 47800 tensor(5.2505, device='cuda:0', grad_fn=<NllLossBackward0>) 47900 tensor(5.2392, device='cuda:0', grad_fn=<NllLossBackward0>) 48000 tensor(4.9209, device='cuda:0', grad_fn=<NllLossBackward0>) 172000 48100 tensor(5.3051, device='cuda:0', grad_fn=<NllLossBackward0>) 48200 tensor(5.1233, device='cuda:0', grad_fn=<NllLossBackward0>) 48300 tensor(5.2021, device='cuda:0', grad_fn=<NllLossBackward0>) 173000 48400 tensor(4.8006, device='cuda:0', grad_fn=<NllLossBackward0>) 48500 tensor(5.0496, device='cuda:0', grad_fn=<NllLossBackward0>) 48600 tensor(5.1875, device='cuda:0', grad_fn=<NllLossBackward0>) 174000 48700 tensor(5.1499, device='cuda:0', grad_fn=<NllLossBackward0>) 48800 tensor(4.9676, device='cuda:0', grad_fn=<NllLossBackward0>) 175000 48900 tensor(5.2299, device='cuda:0', grad_fn=<NllLossBackward0>) 49000 tensor(5.0078, device='cuda:0', grad_fn=<NllLossBackward0>) 49100 tensor(5.1948, device='cuda:0', grad_fn=<NllLossBackward0>) 176000 49200 tensor(5.2544, device='cuda:0', grad_fn=<NllLossBackward0>) 49300 tensor(5.0308, device='cuda:0', grad_fn=<NllLossBackward0>) 49400 tensor(5.2188, device='cuda:0', grad_fn=<NllLossBackward0>) 177000 49500 tensor(5.3279, device='cuda:0', grad_fn=<NllLossBackward0>) 49600 tensor(4.9758, device='cuda:0', grad_fn=<NllLossBackward0>) 49700 tensor(5.0799, device='cuda:0', grad_fn=<NllLossBackward0>) 178000 49800 tensor(5.0324, device='cuda:0', grad_fn=<NllLossBackward0>) 49900 tensor(5.2124, device='cuda:0', grad_fn=<NllLossBackward0>) 50000 tensor(5.1213, device='cuda:0', grad_fn=<NllLossBackward0>) 179000 50100 tensor(5.2672, device='cuda:0', grad_fn=<NllLossBackward0>) 50200 tensor(4.9766, device='cuda:0', grad_fn=<NllLossBackward0>) 180000 50300 tensor(5.2499, device='cuda:0', grad_fn=<NllLossBackward0>) 50400 tensor(4.9455, device='cuda:0', grad_fn=<NllLossBackward0>) 50500 tensor(5.2477, device='cuda:0', grad_fn=<NllLossBackward0>) 181000 50600 tensor(5.0518, device='cuda:0', grad_fn=<NllLossBackward0>) 50700 tensor(4.9514, device='cuda:0', grad_fn=<NllLossBackward0>) 50800 tensor(4.8579, device='cuda:0', grad_fn=<NllLossBackward0>) 182000 50900 tensor(5.1602, device='cuda:0', grad_fn=<NllLossBackward0>) 51000 tensor(5.2193, device='cuda:0', grad_fn=<NllLossBackward0>) 51100 tensor(5.0982, device='cuda:0', grad_fn=<NllLossBackward0>) 183000 51200 tensor(5.2880, device='cuda:0', grad_fn=<NllLossBackward0>) 51300 tensor(4.8375, device='cuda:0', grad_fn=<NllLossBackward0>) 51400 tensor(5.1995, device='cuda:0', grad_fn=<NllLossBackward0>) 184000 51500 tensor(5.1672, device='cuda:0', grad_fn=<NllLossBackward0>) 51600 tensor(5.2276, device='cuda:0', grad_fn=<NllLossBackward0>) 185000 51700 tensor(5.0882, device='cuda:0', grad_fn=<NllLossBackward0>) 51800 tensor(5.2304, device='cuda:0', grad_fn=<NllLossBackward0>) 51900 tensor(5.0226, device='cuda:0', grad_fn=<NllLossBackward0>) 186000 52000 tensor(5.1911, device='cuda:0', grad_fn=<NllLossBackward0>) 52100 tensor(5.3534, device='cuda:0', grad_fn=<NllLossBackward0>) 52200 tensor(5.1108, device='cuda:0', grad_fn=<NllLossBackward0>) 187000 52300 tensor(5.0774, device='cuda:0', grad_fn=<NllLossBackward0>) 52400 tensor(5.1188, device='cuda:0', grad_fn=<NllLossBackward0>) 52500 tensor(5.1506, device='cuda:0', grad_fn=<NllLossBackward0>) 188000 52600 tensor(5.0832, device='cuda:0', grad_fn=<NllLossBackward0>) 52700 tensor(5.2014, device='cuda:0', grad_fn=<NllLossBackward0>) 189000 52800 tensor(5.2696, device='cuda:0', grad_fn=<NllLossBackward0>) 52900 tensor(5.0779, device='cuda:0', grad_fn=<NllLossBackward0>) 53000 tensor(5.0554, device='cuda:0', grad_fn=<NllLossBackward0>) 190000 53100 tensor(5.0068, device='cuda:0', grad_fn=<NllLossBackward0>) 53200 tensor(4.7833, device='cuda:0', grad_fn=<NllLossBackward0>) 53300 tensor(5.2709, device='cuda:0', grad_fn=<NllLossBackward0>) 191000 53400 tensor(5.1926, device='cuda:0', grad_fn=<NllLossBackward0>) 53500 tensor(5.0873, device='cuda:0', grad_fn=<NllLossBackward0>) 53600 tensor(5.1936, device='cuda:0', grad_fn=<NllLossBackward0>) 192000 53700 tensor(5.1878, device='cuda:0', grad_fn=<NllLossBackward0>) 53800 tensor(4.8307, device='cuda:0', grad_fn=<NllLossBackward0>) 53900 tensor(4.8449, device='cuda:0', grad_fn=<NllLossBackward0>) 193000 54000 tensor(5.3690, device='cuda:0', grad_fn=<NllLossBackward0>) 54100 tensor(4.4637, device='cuda:0', grad_fn=<NllLossBackward0>) 194000 54200 tensor(5.2497, device='cuda:0', grad_fn=<NllLossBackward0>) 54300 tensor(5.0490, device='cuda:0', grad_fn=<NllLossBackward0>) 54400 tensor(4.7058, device='cuda:0', grad_fn=<NllLossBackward0>) 195000 54500 tensor(5.2924, device='cuda:0', grad_fn=<NllLossBackward0>) 54600 tensor(5.3235, device='cuda:0', grad_fn=<NllLossBackward0>) 54700 tensor(5.0440, device='cuda:0', grad_fn=<NllLossBackward0>) 196000 54800 tensor(5.2241, device='cuda:0', grad_fn=<NllLossBackward0>) 54900 tensor(4.9492, device='cuda:0', grad_fn=<NllLossBackward0>) 55000 tensor(5.0754, device='cuda:0', grad_fn=<NllLossBackward0>) 197000 55100 tensor(5.1074, device='cuda:0', grad_fn=<NllLossBackward0>) 55200 tensor(5.2155, device='cuda:0', grad_fn=<NllLossBackward0>) 55300 tensor(5.1454, device='cuda:0', grad_fn=<NllLossBackward0>) 198000 55400 tensor(5.2931, device='cuda:0', grad_fn=<NllLossBackward0>) 55500 tensor(5.3043, device='cuda:0', grad_fn=<NllLossBackward0>) 199000 55600 tensor(4.8684, device='cuda:0', grad_fn=<NllLossBackward0>) 55700 tensor(5.0869, device='cuda:0', grad_fn=<NllLossBackward0>) 55800 tensor(4.6445, device='cuda:0', grad_fn=<NllLossBackward0>) 200000 55900 tensor(5.1047, device='cuda:0', grad_fn=<NllLossBackward0>) 56000 tensor(4.9902, device='cuda:0', grad_fn=<NllLossBackward0>) 56100 tensor(5.2816, device='cuda:0', grad_fn=<NllLossBackward0>) 201000 56200 tensor(5.3301, device='cuda:0', grad_fn=<NllLossBackward0>) 56300 tensor(5.1437, device='cuda:0', grad_fn=<NllLossBackward0>) 56400 tensor(5.2254, device='cuda:0', grad_fn=<NllLossBackward0>) 202000 56500 tensor(5.2339, device='cuda:0', grad_fn=<NllLossBackward0>) 56600 tensor(4.9340, device='cuda:0', grad_fn=<NllLossBackward0>) 56700 tensor(5.0758, device='cuda:0', grad_fn=<NllLossBackward0>) 203000 56800 tensor(5.5707, device='cuda:0', grad_fn=<NllLossBackward0>) 56900 tensor(4.9705, device='cuda:0', grad_fn=<NllLossBackward0>) 204000 57000 tensor(5.3555, device='cuda:0', grad_fn=<NllLossBackward0>) 57100 tensor(5.1206, device='cuda:0', grad_fn=<NllLossBackward0>) 57200 tensor(4.9509, device='cuda:0', grad_fn=<NllLossBackward0>) 205000 57300 tensor(5.0644, device='cuda:0', grad_fn=<NllLossBackward0>) 57400 tensor(5.2678, device='cuda:0', grad_fn=<NllLossBackward0>) 57500 tensor(4.9110, device='cuda:0', grad_fn=<NllLossBackward0>) 206000 57600 tensor(4.7788, device='cuda:0', grad_fn=<NllLossBackward0>) 57700 tensor(4.9975, device='cuda:0', grad_fn=<NllLossBackward0>) 57800 tensor(5.3311, device='cuda:0', grad_fn=<NllLossBackward0>) 207000 57900 tensor(5.2600, device='cuda:0', grad_fn=<NllLossBackward0>) 58000 tensor(5.0070, device='cuda:0', grad_fn=<NllLossBackward0>) 208000 58100 tensor(5.3885, device='cuda:0', grad_fn=<NllLossBackward0>) 58200 tensor(5.4209, device='cuda:0', grad_fn=<NllLossBackward0>) 58300 tensor(4.9688, device='cuda:0', grad_fn=<NllLossBackward0>) 209000 58400 tensor(4.8874, device='cuda:0', grad_fn=<NllLossBackward0>) 58500 tensor(4.9976, device='cuda:0', grad_fn=<NllLossBackward0>) 58600 tensor(5.4753, device='cuda:0', grad_fn=<NllLossBackward0>) 210000 58700 tensor(4.9572, device='cuda:0', grad_fn=<NllLossBackward0>) 58800 tensor(5.1014, device='cuda:0', grad_fn=<NllLossBackward0>) 58900 tensor(5.2920, device='cuda:0', grad_fn=<NllLossBackward0>) 211000 59000 tensor(5.2007, device='cuda:0', grad_fn=<NllLossBackward0>) 59100 tensor(5.0399, device='cuda:0', grad_fn=<NllLossBackward0>) 59200 tensor(5.1078, device='cuda:0', grad_fn=<NllLossBackward0>) 212000 59300 tensor(4.9478, device='cuda:0', grad_fn=<NllLossBackward0>) 59400 tensor(5.1439, device='cuda:0', grad_fn=<NllLossBackward0>) 213000 59500 tensor(5.0825, device='cuda:0', grad_fn=<NllLossBackward0>) 59600 tensor(5.0066, device='cuda:0', grad_fn=<NllLossBackward0>) 59700 tensor(4.8766, device='cuda:0', grad_fn=<NllLossBackward0>) 214000 59800 tensor(5.0725, device='cuda:0', grad_fn=<NllLossBackward0>) 59900 tensor(5.1101, device='cuda:0', grad_fn=<NllLossBackward0>) 60000 tensor(5.1072, device='cuda:0', grad_fn=<NllLossBackward0>) 215000 60100 tensor(5.1821, device='cuda:0', grad_fn=<NllLossBackward0>) 60200 tensor(5.2384, device='cuda:0', grad_fn=<NllLossBackward0>) 60300 tensor(5.0115, device='cuda:0', grad_fn=<NllLossBackward0>) 216000 60400 tensor(5.1308, device='cuda:0', grad_fn=<NllLossBackward0>) 60500 tensor(5.1859, device='cuda:0', grad_fn=<NllLossBackward0>) 60600 tensor(5.0918, device='cuda:0', grad_fn=<NllLossBackward0>) 217000 60700 tensor(5.0108, device='cuda:0', grad_fn=<NllLossBackward0>) 60800 tensor(5.1547, device='cuda:0', grad_fn=<NllLossBackward0>) 218000 60900 tensor(5.2653, device='cuda:0', grad_fn=<NllLossBackward0>) 61000 tensor(5.2235, device='cuda:0', grad_fn=<NllLossBackward0>) 61100 tensor(5.1075, device='cuda:0', grad_fn=<NllLossBackward0>) 219000 61200 tensor(4.8467, device='cuda:0', grad_fn=<NllLossBackward0>) 61300 tensor(5.0838, device='cuda:0', grad_fn=<NllLossBackward0>) 61400 tensor(5.2590, device='cuda:0', grad_fn=<NllLossBackward0>) 220000 61500 tensor(5.1743, device='cuda:0', grad_fn=<NllLossBackward0>) 61600 tensor(4.9954, device='cuda:0', grad_fn=<NllLossBackward0>) 61700 tensor(4.9919, device='cuda:0', grad_fn=<NllLossBackward0>) 221000 61800 tensor(5.0254, device='cuda:0', grad_fn=<NllLossBackward0>) 61900 tensor(5.0908, device='cuda:0', grad_fn=<NllLossBackward0>) 62000 tensor(5.2753, device='cuda:0', grad_fn=<NllLossBackward0>) 222000 62100 tensor(4.8669, device='cuda:0', grad_fn=<NllLossBackward0>) 62200 tensor(4.9641, device='cuda:0', grad_fn=<NllLossBackward0>) 223000 62300 tensor(5.0582, device='cuda:0', grad_fn=<NllLossBackward0>) 62400 tensor(4.9853, device='cuda:0', grad_fn=<NllLossBackward0>) 62500 tensor(5.0922, device='cuda:0', grad_fn=<NllLossBackward0>) 224000 62600 tensor(4.9824, device='cuda:0', grad_fn=<NllLossBackward0>) 62700 tensor(5.1782, device='cuda:0', grad_fn=<NllLossBackward0>) 62800 tensor(4.8954, device='cuda:0', grad_fn=<NllLossBackward0>) 225000 62900 tensor(5.3762, device='cuda:0', grad_fn=<NllLossBackward0>) 63000 tensor(5.2436, device='cuda:0', grad_fn=<NllLossBackward0>) 63100 tensor(5.3535, device='cuda:0', grad_fn=<NllLossBackward0>) 226000 63200 tensor(5.1772, device='cuda:0', grad_fn=<NllLossBackward0>) 63300 tensor(4.9112, device='cuda:0', grad_fn=<NllLossBackward0>) 63400 tensor(5.0962, device='cuda:0', grad_fn=<NllLossBackward0>) 227000 63500 tensor(4.9657, device='cuda:0', grad_fn=<NllLossBackward0>) 63600 tensor(5.0014, device='cuda:0', grad_fn=<NllLossBackward0>) 228000 63700 tensor(5.0686, device='cuda:0', grad_fn=<NllLossBackward0>) 63800 tensor(5.3235, device='cuda:0', grad_fn=<NllLossBackward0>) 63900 tensor(5.3906, device='cuda:0', grad_fn=<NllLossBackward0>) 229000 64000 tensor(4.9716, device='cuda:0', grad_fn=<NllLossBackward0>) 64100 tensor(5.2290, device='cuda:0', grad_fn=<NllLossBackward0>) 64200 tensor(5.1423, device='cuda:0', grad_fn=<NllLossBackward0>) 230000 64300 tensor(5.1310, device='cuda:0', grad_fn=<NllLossBackward0>) 64400 tensor(4.9695, device='cuda:0', grad_fn=<NllLossBackward0>) 64500 tensor(4.9604, device='cuda:0', grad_fn=<NllLossBackward0>) 231000 64600 tensor(4.9843, device='cuda:0', grad_fn=<NllLossBackward0>) 64700 tensor(5.4459, device='cuda:0', grad_fn=<NllLossBackward0>) 64800 tensor(5.1474, device='cuda:0', grad_fn=<NllLossBackward0>) 232000 64900 tensor(5.2715, device='cuda:0', grad_fn=<NllLossBackward0>) 65000 tensor(5.2314, device='cuda:0', grad_fn=<NllLossBackward0>) 233000 65100 tensor(4.7774, device='cuda:0', grad_fn=<NllLossBackward0>) 65200 tensor(5.3146, device='cuda:0', grad_fn=<NllLossBackward0>) 65300 tensor(5.1604, device='cuda:0', grad_fn=<NllLossBackward0>) 234000 65400 tensor(5.0736, device='cuda:0', grad_fn=<NllLossBackward0>) 65500 tensor(5.1152, device='cuda:0', grad_fn=<NllLossBackward0>) 65600 tensor(5.1409, device='cuda:0', grad_fn=<NllLossBackward0>) 235000 65700 tensor(4.9068, device='cuda:0', grad_fn=<NllLossBackward0>) 65800 tensor(4.9353, device='cuda:0', grad_fn=<NllLossBackward0>) 65900 tensor(5.1612, device='cuda:0', grad_fn=<NllLossBackward0>) 236000 66000 tensor(5.1866, device='cuda:0', grad_fn=<NllLossBackward0>) 66100 tensor(4.8965, device='cuda:0', grad_fn=<NllLossBackward0>) 237000 66200 tensor(4.9215, device='cuda:0', grad_fn=<NllLossBackward0>) 66300 tensor(4.8641, device='cuda:0', grad_fn=<NllLossBackward0>) 66400 tensor(5.1550, device='cuda:0', grad_fn=<NllLossBackward0>) 238000 66500 tensor(4.8722, device='cuda:0', grad_fn=<NllLossBackward0>) 66600 tensor(5.0268, device='cuda:0', grad_fn=<NllLossBackward0>) 66700 tensor(4.9948, device='cuda:0', grad_fn=<NllLossBackward0>) 239000 66800 tensor(5.0565, device='cuda:0', grad_fn=<NllLossBackward0>) 66900 tensor(5.2030, device='cuda:0', grad_fn=<NllLossBackward0>) 67000 tensor(5.2199, device='cuda:0', grad_fn=<NllLossBackward0>) 240000 67100 tensor(5.0560, device='cuda:0', grad_fn=<NllLossBackward0>) 67200 tensor(5.4027, device='cuda:0', grad_fn=<NllLossBackward0>) 67300 tensor(5.0631, device='cuda:0', grad_fn=<NllLossBackward0>) 241000 67400 tensor(5.2687, device='cuda:0', grad_fn=<NllLossBackward0>) 67500 tensor(5.0728, device='cuda:0', grad_fn=<NllLossBackward0>) 242000 67600 tensor(5.1050, device='cuda:0', grad_fn=<NllLossBackward0>) 67700 tensor(5.0495, device='cuda:0', grad_fn=<NllLossBackward0>) 67800 tensor(5.2014, device='cuda:0', grad_fn=<NllLossBackward0>) 243000 67900 tensor(4.9060, device='cuda:0', grad_fn=<NllLossBackward0>) 68000 tensor(4.9703, device='cuda:0', grad_fn=<NllLossBackward0>) 68100 tensor(4.8457, device='cuda:0', grad_fn=<NllLossBackward0>) 244000 68200 tensor(5.1298, device='cuda:0', grad_fn=<NllLossBackward0>) 68300 tensor(5.0272, device='cuda:0', grad_fn=<NllLossBackward0>) 68400 tensor(5.1238, device='cuda:0', grad_fn=<NllLossBackward0>) 245000 68500 tensor(5.1778, device='cuda:0', grad_fn=<NllLossBackward0>) 68600 tensor(5.0997, device='cuda:0', grad_fn=<NllLossBackward0>) 68700 tensor(4.7577, device='cuda:0', grad_fn=<NllLossBackward0>) 246000 68800 tensor(5.1663, device='cuda:0', grad_fn=<NllLossBackward0>) 68900 tensor(4.8385, device='cuda:0', grad_fn=<NllLossBackward0>) 247000 69000 tensor(5.0810, device='cuda:0', grad_fn=<NllLossBackward0>) 69100 tensor(4.8565, device='cuda:0', grad_fn=<NllLossBackward0>) 69200 tensor(4.8088, device='cuda:0', grad_fn=<NllLossBackward0>) 248000 69300 tensor(5.2732, device='cuda:0', grad_fn=<NllLossBackward0>) 69400 tensor(5.1549, device='cuda:0', grad_fn=<NllLossBackward0>) 69500 tensor(5.1323, device='cuda:0', grad_fn=<NllLossBackward0>) 249000 69600 tensor(5.0379, device='cuda:0', grad_fn=<NllLossBackward0>) 69700 tensor(5.3768, device='cuda:0', grad_fn=<NllLossBackward0>) 69800 tensor(4.9056, device='cuda:0', grad_fn=<NllLossBackward0>) 250000 69900 tensor(5.4631, device='cuda:0', grad_fn=<NllLossBackward0>) 70000 tensor(5.1071, device='cuda:0', grad_fn=<NllLossBackward0>) 70100 tensor(5.0137, device='cuda:0', grad_fn=<NllLossBackward0>) 251000 70200 tensor(5.1769, device='cuda:0', grad_fn=<NllLossBackward0>) 70300 tensor(4.8435, device='cuda:0', grad_fn=<NllLossBackward0>) 252000 70400 tensor(5.2020, device='cuda:0', grad_fn=<NllLossBackward0>) 70500 tensor(5.1395, device='cuda:0', grad_fn=<NllLossBackward0>) 70600 tensor(5.3541, device='cuda:0', grad_fn=<NllLossBackward0>) 253000 70700 tensor(5.3092, device='cuda:0', grad_fn=<NllLossBackward0>) 70800 tensor(5.4793, device='cuda:0', grad_fn=<NllLossBackward0>) 70900 tensor(5.2799, device='cuda:0', grad_fn=<NllLossBackward0>) 254000 71000 tensor(5.0872, device='cuda:0', grad_fn=<NllLossBackward0>) 71100 tensor(5.1728, device='cuda:0', grad_fn=<NllLossBackward0>) 71200 tensor(5.2739, device='cuda:0', grad_fn=<NllLossBackward0>) 255000 71300 tensor(4.8960, device='cuda:0', grad_fn=<NllLossBackward0>) 71400 tensor(5.0185, device='cuda:0', grad_fn=<NllLossBackward0>) 71500 tensor(5.0367, device='cuda:0', grad_fn=<NllLossBackward0>) 256000 71600 tensor(5.1992, device='cuda:0', grad_fn=<NllLossBackward0>) 71700 tensor(5.2508, device='cuda:0', grad_fn=<NllLossBackward0>) 257000 71800 tensor(5.3864, device='cuda:0', grad_fn=<NllLossBackward0>) 71900 tensor(5.0381, device='cuda:0', grad_fn=<NllLossBackward0>) 72000 tensor(5.4758, device='cuda:0', grad_fn=<NllLossBackward0>) 258000 72100 tensor(5.1863, device='cuda:0', grad_fn=<NllLossBackward0>) 72200 tensor(5.0258, device='cuda:0', grad_fn=<NllLossBackward0>) 72300 tensor(5.1253, device='cuda:0', grad_fn=<NllLossBackward0>) 259000 72400 tensor(5.0412, device='cuda:0', grad_fn=<NllLossBackward0>) 72500 tensor(5.1304, device='cuda:0', grad_fn=<NllLossBackward0>) 72600 tensor(5.1833, device='cuda:0', grad_fn=<NllLossBackward0>) 260000 72700 tensor(4.8666, device='cuda:0', grad_fn=<NllLossBackward0>) 72800 tensor(5.3088, device='cuda:0', grad_fn=<NllLossBackward0>) 72900 tensor(5.0482, device='cuda:0', grad_fn=<NllLossBackward0>) 261000 73000 tensor(5.2591, device='cuda:0', grad_fn=<NllLossBackward0>) 73100 tensor(4.9459, device='cuda:0', grad_fn=<NllLossBackward0>) 262000 73200 tensor(5.0086, device='cuda:0', grad_fn=<NllLossBackward0>) 73300 tensor(4.9859, device='cuda:0', grad_fn=<NllLossBackward0>) 73400 tensor(5.2442, device='cuda:0', grad_fn=<NllLossBackward0>) 263000 73500 tensor(5.0674, device='cuda:0', grad_fn=<NllLossBackward0>) 73600 tensor(5.0003, device='cuda:0', grad_fn=<NllLossBackward0>) 73700 tensor(5.1476, device='cuda:0', grad_fn=<NllLossBackward0>) 264000 73800 tensor(5.1845, device='cuda:0', grad_fn=<NllLossBackward0>) 73900 tensor(5.1606, device='cuda:0', grad_fn=<NllLossBackward0>) 74000 tensor(5.3307, device='cuda:0', grad_fn=<NllLossBackward0>) 265000 74100 tensor(5.0139, device='cuda:0', grad_fn=<NllLossBackward0>) 74200 tensor(5.0479, device='cuda:0', grad_fn=<NllLossBackward0>) 74300 tensor(5.0041, device='cuda:0', grad_fn=<NllLossBackward0>) 266000 74400 tensor(5.1568, device='cuda:0', grad_fn=<NllLossBackward0>) 74500 tensor(4.7452, device='cuda:0', grad_fn=<NllLossBackward0>) 267000 74600 tensor(5.2662, device='cuda:0', grad_fn=<NllLossBackward0>) 74700 tensor(5.3145, device='cuda:0', grad_fn=<NllLossBackward0>) 74800 tensor(5.0200, device='cuda:0', grad_fn=<NllLossBackward0>) 268000 74900 tensor(4.8437, device='cuda:0', grad_fn=<NllLossBackward0>) 75000 tensor(5.1458, device='cuda:0', grad_fn=<NllLossBackward0>) 75100 tensor(5.3639, device='cuda:0', grad_fn=<NllLossBackward0>) 269000 75200 tensor(5.0922, device='cuda:0', grad_fn=<NllLossBackward0>) 75300 tensor(5.3002, device='cuda:0', grad_fn=<NllLossBackward0>) 75400 tensor(5.0810, device='cuda:0', grad_fn=<NllLossBackward0>) 270000 75500 tensor(5.3591, device='cuda:0', grad_fn=<NllLossBackward0>) 75600 tensor(5.0206, device='cuda:0', grad_fn=<NllLossBackward0>) 271000 75700 tensor(5.0759, device='cuda:0', grad_fn=<NllLossBackward0>) 75800 tensor(5.0702, device='cuda:0', grad_fn=<NllLossBackward0>) 75900 tensor(5.1349, device='cuda:0', grad_fn=<NllLossBackward0>) 272000 76000 tensor(4.8964, device='cuda:0', grad_fn=<NllLossBackward0>) 76100 tensor(4.8618, device='cuda:0', grad_fn=<NllLossBackward0>) 76200 tensor(4.9370, device='cuda:0', grad_fn=<NllLossBackward0>) 273000 76300 tensor(5.0666, device='cuda:0', grad_fn=<NllLossBackward0>) 76400 tensor(4.9261, device='cuda:0', grad_fn=<NllLossBackward0>) 76500 tensor(5.0910, device='cuda:0', grad_fn=<NllLossBackward0>) 274000 76600 tensor(5.1985, device='cuda:0', grad_fn=<NllLossBackward0>) 76700 tensor(4.8590, device='cuda:0', grad_fn=<NllLossBackward0>) 76800 tensor(4.7918, device='cuda:0', grad_fn=<NllLossBackward0>) 275000 76900 tensor(5.1386, device='cuda:0', grad_fn=<NllLossBackward0>) 77000 tensor(5.0233, device='cuda:0', grad_fn=<NllLossBackward0>) 276000 77100 tensor(5.3827, device='cuda:0', grad_fn=<NllLossBackward0>) 77200 tensor(5.0204, device='cuda:0', grad_fn=<NllLossBackward0>) 77300 tensor(4.9805, device='cuda:0', grad_fn=<NllLossBackward0>) 277000 77400 tensor(5.0943, device='cuda:0', grad_fn=<NllLossBackward0>) 77500 tensor(4.7828, device='cuda:0', grad_fn=<NllLossBackward0>) 77600 tensor(5.1103, device='cuda:0', grad_fn=<NllLossBackward0>) 278000 77700 tensor(5.2850, device='cuda:0', grad_fn=<NllLossBackward0>) 77800 tensor(4.8699, device='cuda:0', grad_fn=<NllLossBackward0>) 77900 tensor(4.7829, device='cuda:0', grad_fn=<NllLossBackward0>) 279000 78000 tensor(5.2236, device='cuda:0', grad_fn=<NllLossBackward0>) 78100 tensor(4.8891, device='cuda:0', grad_fn=<NllLossBackward0>) 78200 tensor(5.1773, device='cuda:0', grad_fn=<NllLossBackward0>) 280000 78300 tensor(5.0470, device='cuda:0', grad_fn=<NllLossBackward0>) 78400 tensor(5.0570, device='cuda:0', grad_fn=<NllLossBackward0>) 281000 78500 tensor(5.4172, device='cuda:0', grad_fn=<NllLossBackward0>) 78600 tensor(4.8807, device='cuda:0', grad_fn=<NllLossBackward0>) 78700 tensor(5.3329, device='cuda:0', grad_fn=<NllLossBackward0>) 282000 78800 tensor(5.2173, device='cuda:0', grad_fn=<NllLossBackward0>) 78900 tensor(5.4813, device='cuda:0', grad_fn=<NllLossBackward0>) 79000 tensor(4.9973, device='cuda:0', grad_fn=<NllLossBackward0>) 283000 79100 tensor(4.7734, device='cuda:0', grad_fn=<NllLossBackward0>) 79200 tensor(4.8357, device='cuda:0', grad_fn=<NllLossBackward0>) 79300 tensor(5.3520, device='cuda:0', grad_fn=<NllLossBackward0>) 284000 79400 tensor(5.1467, device='cuda:0', grad_fn=<NllLossBackward0>) 79500 tensor(5.0294, device='cuda:0', grad_fn=<NllLossBackward0>) 79600 tensor(5.6394, device='cuda:0', grad_fn=<NllLossBackward0>) 285000 79700 tensor(4.7860, device='cuda:0', grad_fn=<NllLossBackward0>) 79800 tensor(5.0916, device='cuda:0', grad_fn=<NllLossBackward0>) 286000 79900 tensor(5.0640, device='cuda:0', grad_fn=<NllLossBackward0>) 80000 tensor(5.2131, device='cuda:0', grad_fn=<NllLossBackward0>) 80100 tensor(5.2826, device='cuda:0', grad_fn=<NllLossBackward0>) 287000 80200 tensor(5.3061, device='cuda:0', grad_fn=<NllLossBackward0>) 80300 tensor(5.2656, device='cuda:0', grad_fn=<NllLossBackward0>) 80400 tensor(5.3741, device='cuda:0', grad_fn=<NllLossBackward0>) 288000 80500 tensor(5.0930, device='cuda:0', grad_fn=<NllLossBackward0>) 80600 tensor(5.2616, device='cuda:0', grad_fn=<NllLossBackward0>) 80700 tensor(5.3157, device='cuda:0', grad_fn=<NllLossBackward0>) 289000 80800 tensor(4.8360, device='cuda:0', grad_fn=<NllLossBackward0>) 80900 tensor(5.2270, device='cuda:0', grad_fn=<NllLossBackward0>) 290000 81000 tensor(5.1687, device='cuda:0', grad_fn=<NllLossBackward0>) 81100 tensor(5.1558, device='cuda:0', grad_fn=<NllLossBackward0>) 81200 tensor(4.9090, device='cuda:0', grad_fn=<NllLossBackward0>) 291000 81300 tensor(4.9953, device='cuda:0', grad_fn=<NllLossBackward0>) 81400 tensor(5.1353, device='cuda:0', grad_fn=<NllLossBackward0>) 81500 tensor(5.1045, device='cuda:0', grad_fn=<NllLossBackward0>) 292000 81600 tensor(5.1020, device='cuda:0', grad_fn=<NllLossBackward0>) 81700 tensor(5.2123, device='cuda:0', grad_fn=<NllLossBackward0>) 81800 tensor(5.0393, device='cuda:0', grad_fn=<NllLossBackward0>) 293000 81900 tensor(5.2712, device='cuda:0', grad_fn=<NllLossBackward0>) 82000 tensor(5.0969, device='cuda:0', grad_fn=<NllLossBackward0>) 82100 tensor(5.1216, device='cuda:0', grad_fn=<NllLossBackward0>) 294000 82200 tensor(5.0557, device='cuda:0', grad_fn=<NllLossBackward0>) 82300 tensor(4.9278, device='cuda:0', grad_fn=<NllLossBackward0>) 295000 82400 tensor(5.1703, device='cuda:0', grad_fn=<NllLossBackward0>) 82500 tensor(4.8852, device='cuda:0', grad_fn=<NllLossBackward0>) 82600 tensor(5.0861, device='cuda:0', grad_fn=<NllLossBackward0>) 296000 82700 tensor(5.1290, device='cuda:0', grad_fn=<NllLossBackward0>) 82800 tensor(5.1089, device='cuda:0', grad_fn=<NllLossBackward0>) 82900 tensor(4.7936, device='cuda:0', grad_fn=<NllLossBackward0>) 297000 83000 tensor(5.1975, device='cuda:0', grad_fn=<NllLossBackward0>) 83100 tensor(5.0460, device='cuda:0', grad_fn=<NllLossBackward0>) 83200 tensor(5.2952, device='cuda:0', grad_fn=<NllLossBackward0>) 298000 83300 tensor(4.9863, device='cuda:0', grad_fn=<NllLossBackward0>) 83400 tensor(5.0724, device='cuda:0', grad_fn=<NllLossBackward0>) 83500 tensor(5.0200, device='cuda:0', grad_fn=<NllLossBackward0>) 299000 83600 tensor(4.9496, device='cuda:0', grad_fn=<NllLossBackward0>) 83700 tensor(5.1706, device='cuda:0', grad_fn=<NllLossBackward0>) 300000 83800 tensor(4.8822, device='cuda:0', grad_fn=<NllLossBackward0>) 83900 tensor(4.9673, device='cuda:0', grad_fn=<NllLossBackward0>) 84000 tensor(5.1188, device='cuda:0', grad_fn=<NllLossBackward0>) 301000 84100 tensor(5.4809, device='cuda:0', grad_fn=<NllLossBackward0>) 84200 tensor(4.6303, device='cuda:0', grad_fn=<NllLossBackward0>) 84300 tensor(5.0608, device='cuda:0', grad_fn=<NllLossBackward0>) 302000 84400 tensor(4.8835, device='cuda:0', grad_fn=<NllLossBackward0>) 84500 tensor(5.0710, device='cuda:0', grad_fn=<NllLossBackward0>) 84600 tensor(5.2347, device='cuda:0', grad_fn=<NllLossBackward0>) 303000 84700 tensor(4.9578, device='cuda:0', grad_fn=<NllLossBackward0>) 84800 tensor(4.9840, device='cuda:0', grad_fn=<NllLossBackward0>) 84900 tensor(5.2946, device='cuda:0', grad_fn=<NllLossBackward0>) 304000 85000 tensor(5.1757, device='cuda:0', grad_fn=<NllLossBackward0>) 85100 tensor(5.0449, device='cuda:0', grad_fn=<NllLossBackward0>) 305000 85200 tensor(5.0524, device='cuda:0', grad_fn=<NllLossBackward0>) 85300 tensor(5.3156, device='cuda:0', grad_fn=<NllLossBackward0>) 85400 tensor(5.2982, device='cuda:0', grad_fn=<NllLossBackward0>) 306000 85500 tensor(4.9904, device='cuda:0', grad_fn=<NllLossBackward0>) 85600 tensor(5.1111, device='cuda:0', grad_fn=<NllLossBackward0>) 85700 tensor(5.1132, device='cuda:0', grad_fn=<NllLossBackward0>) 307000 85800 tensor(5.1134, device='cuda:0', grad_fn=<NllLossBackward0>) 85900 tensor(5.3567, device='cuda:0', grad_fn=<NllLossBackward0>) 86000 tensor(5.1374, device='cuda:0', grad_fn=<NllLossBackward0>) 308000 86100 tensor(4.8926, device='cuda:0', grad_fn=<NllLossBackward0>) 86200 tensor(5.0359, device='cuda:0', grad_fn=<NllLossBackward0>) 86300 tensor(5.0061, device='cuda:0', grad_fn=<NllLossBackward0>) 309000 86400 tensor(5.0968, device='cuda:0', grad_fn=<NllLossBackward0>) 86500 tensor(5.1935, device='cuda:0', grad_fn=<NllLossBackward0>) 310000 86600 tensor(5.0132, device='cuda:0', grad_fn=<NllLossBackward0>) 86700 tensor(4.9140, device='cuda:0', grad_fn=<NllLossBackward0>) 86800 tensor(4.9166, device='cuda:0', grad_fn=<NllLossBackward0>) 311000 86900 tensor(4.8179, device='cuda:0', grad_fn=<NllLossBackward0>) 87000 tensor(5.3136, device='cuda:0', grad_fn=<NllLossBackward0>) 87100 tensor(5.1485, device='cuda:0', grad_fn=<NllLossBackward0>) 312000 87200 tensor(5.1635, device='cuda:0', grad_fn=<NllLossBackward0>) 87300 tensor(5.0104, device='cuda:0', grad_fn=<NllLossBackward0>) 87400 tensor(5.0568, device='cuda:0', grad_fn=<NllLossBackward0>) 313000 87500 tensor(5.1699, device='cuda:0', grad_fn=<NllLossBackward0>) 87600 tensor(5.2978, device='cuda:0', grad_fn=<NllLossBackward0>) 87700 tensor(4.9475, device='cuda:0', grad_fn=<NllLossBackward0>) 314000 87800 tensor(4.8997, device='cuda:0', grad_fn=<NllLossBackward0>) 87900 tensor(5.1651, device='cuda:0', grad_fn=<NllLossBackward0>) 315000 88000 tensor(4.6506, device='cuda:0', grad_fn=<NllLossBackward0>) 88100 tensor(5.0203, device='cuda:0', grad_fn=<NllLossBackward0>) 88200 tensor(5.2400, device='cuda:0', grad_fn=<NllLossBackward0>) 316000 88300 tensor(5.1478, device='cuda:0', grad_fn=<NllLossBackward0>) 88400 tensor(5.2234, device='cuda:0', grad_fn=<NllLossBackward0>) 88500 tensor(4.8418, device='cuda:0', grad_fn=<NllLossBackward0>) 317000 88600 tensor(4.8617, device='cuda:0', grad_fn=<NllLossBackward0>) 88700 tensor(4.8720, device='cuda:0', grad_fn=<NllLossBackward0>) 88800 tensor(4.8572, device='cuda:0', grad_fn=<NllLossBackward0>) 318000 88900 tensor(4.7781, device='cuda:0', grad_fn=<NllLossBackward0>) 89000 tensor(4.5953, device='cuda:0', grad_fn=<NllLossBackward0>) 89100 tensor(5.1780, device='cuda:0', grad_fn=<NllLossBackward0>) 319000 89200 tensor(4.9773, device='cuda:0', grad_fn=<NllLossBackward0>) 89300 tensor(5.3703, device='cuda:0', grad_fn=<NllLossBackward0>) 320000 89400 tensor(4.7298, device='cuda:0', grad_fn=<NllLossBackward0>) 89500 tensor(5.0713, device='cuda:0', grad_fn=<NllLossBackward0>) 89600 tensor(5.3035, device='cuda:0', grad_fn=<NllLossBackward0>) 321000 89700 tensor(4.8603, device='cuda:0', grad_fn=<NllLossBackward0>) 89800 tensor(5.0780, device='cuda:0', grad_fn=<NllLossBackward0>) 89900 tensor(5.0922, device='cuda:0', grad_fn=<NllLossBackward0>) 322000 90000 tensor(5.0946, device='cuda:0', grad_fn=<NllLossBackward0>) 90100 tensor(4.9839, device='cuda:0', grad_fn=<NllLossBackward0>) 90200 tensor(5.0898, device='cuda:0', grad_fn=<NllLossBackward0>) 323000 90300 tensor(5.0876, device='cuda:0', grad_fn=<NllLossBackward0>) 90400 tensor(4.7618, device='cuda:0', grad_fn=<NllLossBackward0>) 90500 tensor(4.9550, device='cuda:0', grad_fn=<NllLossBackward0>) 324000 90600 tensor(5.2409, device='cuda:0', grad_fn=<NllLossBackward0>) 90700 tensor(4.8453, device='cuda:0', grad_fn=<NllLossBackward0>) 325000 90800 tensor(5.0906, device='cuda:0', grad_fn=<NllLossBackward0>) 90900 tensor(5.1866, device='cuda:0', grad_fn=<NllLossBackward0>) 91000 tensor(5.1850, device='cuda:0', grad_fn=<NllLossBackward0>) 326000 91100 tensor(4.7448, device='cuda:0', grad_fn=<NllLossBackward0>) 91200 tensor(5.2125, device='cuda:0', grad_fn=<NllLossBackward0>) 91300 tensor(5.3846, device='cuda:0', grad_fn=<NllLossBackward0>) 327000 91400 tensor(4.7534, device='cuda:0', grad_fn=<NllLossBackward0>) 91500 tensor(4.9951, device='cuda:0', grad_fn=<NllLossBackward0>) 91600 tensor(5.2443, device='cuda:0', grad_fn=<NllLossBackward0>) 328000 91700 tensor(5.0912, device='cuda:0', grad_fn=<NllLossBackward0>) 91800 tensor(5.2768, device='cuda:0', grad_fn=<NllLossBackward0>) 329000 91900 tensor(5.0739, device='cuda:0', grad_fn=<NllLossBackward0>) 92000 tensor(5.1095, device='cuda:0', grad_fn=<NllLossBackward0>) 92100 tensor(4.9860, device='cuda:0', grad_fn=<NllLossBackward0>) 330000 92200 tensor(4.6185, device='cuda:0', grad_fn=<NllLossBackward0>) 92300 tensor(5.1181, device='cuda:0', grad_fn=<NllLossBackward0>) 92400 tensor(5.1625, device='cuda:0', grad_fn=<NllLossBackward0>) 331000 92500 tensor(5.2200, device='cuda:0', grad_fn=<NllLossBackward0>) 92600 tensor(5.0955, device='cuda:0', grad_fn=<NllLossBackward0>) 92700 tensor(4.8214, device='cuda:0', grad_fn=<NllLossBackward0>) 332000 92800 tensor(4.8953, device='cuda:0', grad_fn=<NllLossBackward0>) 92900 tensor(5.4247, device='cuda:0', grad_fn=<NllLossBackward0>) 93000 tensor(5.1056, device='cuda:0', grad_fn=<NllLossBackward0>) 333000 93100 tensor(5.0311, device='cuda:0', grad_fn=<NllLossBackward0>) 93200 tensor(5.2195, device='cuda:0', grad_fn=<NllLossBackward0>) 334000 93300 tensor(5.1673, device='cuda:0', grad_fn=<NllLossBackward0>) 93400 tensor(5.2275, device='cuda:0', grad_fn=<NllLossBackward0>) 93500 tensor(4.9526, device='cuda:0', grad_fn=<NllLossBackward0>) 335000 93600 tensor(5.0896, device='cuda:0', grad_fn=<NllLossBackward0>) 93700 tensor(5.1565, device='cuda:0', grad_fn=<NllLossBackward0>) 93800 tensor(4.9091, device='cuda:0', grad_fn=<NllLossBackward0>) 336000 93900 tensor(5.1251, device='cuda:0', grad_fn=<NllLossBackward0>) 94000 tensor(5.0433, device='cuda:0', grad_fn=<NllLossBackward0>) 94100 tensor(4.9108, device='cuda:0', grad_fn=<NllLossBackward0>) 337000 94200 tensor(5.1725, device='cuda:0', grad_fn=<NllLossBackward0>) 94300 tensor(5.0600, device='cuda:0', grad_fn=<NllLossBackward0>) 94400 tensor(5.0994, device='cuda:0', grad_fn=<NllLossBackward0>) 338000 94500 tensor(4.9514, device='cuda:0', grad_fn=<NllLossBackward0>) 94600 tensor(5.2426, device='cuda:0', grad_fn=<NllLossBackward0>) 339000 94700 tensor(4.7794, device='cuda:0', grad_fn=<NllLossBackward0>) 94800 tensor(5.2669, device='cuda:0', grad_fn=<NllLossBackward0>) 94900 tensor(5.2703, device='cuda:0', grad_fn=<NllLossBackward0>) 340000 95000 tensor(5.3905, device='cuda:0', grad_fn=<NllLossBackward0>) 95100 tensor(5.0774, device='cuda:0', grad_fn=<NllLossBackward0>) 95200 tensor(4.9328, device='cuda:0', grad_fn=<NllLossBackward0>) 341000 95300 tensor(5.2937, device='cuda:0', grad_fn=<NllLossBackward0>) 95400 tensor(4.7542, device='cuda:0', grad_fn=<NllLossBackward0>) 95500 tensor(4.9919, device='cuda:0', grad_fn=<NllLossBackward0>) 342000 95600 tensor(4.9873, device='cuda:0', grad_fn=<NllLossBackward0>) 95700 tensor(5.1582, device='cuda:0', grad_fn=<NllLossBackward0>) 95800 tensor(5.0927, device='cuda:0', grad_fn=<NllLossBackward0>) 343000 95900 tensor(4.8471, device='cuda:0', grad_fn=<NllLossBackward0>) 96000 tensor(5.1951, device='cuda:0', grad_fn=<NllLossBackward0>) 344000 96100 tensor(4.9727, device='cuda:0', grad_fn=<NllLossBackward0>) 96200 tensor(4.9915, device='cuda:0', grad_fn=<NllLossBackward0>) 96300 tensor(5.0515, device='cuda:0', grad_fn=<NllLossBackward0>) 345000 96400 tensor(5.1081, device='cuda:0', grad_fn=<NllLossBackward0>) 96500 tensor(5.3060, device='cuda:0', grad_fn=<NllLossBackward0>) 96600 tensor(5.0907, device='cuda:0', grad_fn=<NllLossBackward0>) 346000 96700 tensor(5.2215, device='cuda:0', grad_fn=<NllLossBackward0>) 96800 tensor(5.1097, device='cuda:0', grad_fn=<NllLossBackward0>) 96900 tensor(5.2757, device='cuda:0', grad_fn=<NllLossBackward0>) 347000 97000 tensor(4.9539, device='cuda:0', grad_fn=<NllLossBackward0>) 97100 tensor(5.0553, device='cuda:0', grad_fn=<NllLossBackward0>) 97200 tensor(5.1827, device='cuda:0', grad_fn=<NllLossBackward0>) 348000 97300 tensor(5.0123, device='cuda:0', grad_fn=<NllLossBackward0>) 97400 tensor(5.0624, device='cuda:0', grad_fn=<NllLossBackward0>) 349000 97500 tensor(5.2006, device='cuda:0', grad_fn=<NllLossBackward0>) 97600 tensor(4.9950, device='cuda:0', grad_fn=<NllLossBackward0>) 97700 tensor(5.0549, device='cuda:0', grad_fn=<NllLossBackward0>) 350000 97800 tensor(5.0475, device='cuda:0', grad_fn=<NllLossBackward0>) 97900 tensor(4.8108, device='cuda:0', grad_fn=<NllLossBackward0>) 98000 tensor(4.9221, device='cuda:0', grad_fn=<NllLossBackward0>) 351000 98100 tensor(5.2524, device='cuda:0', grad_fn=<NllLossBackward0>) 98200 tensor(4.9745, device='cuda:0', grad_fn=<NllLossBackward0>) 98300 tensor(5.1369, device='cuda:0', grad_fn=<NllLossBackward0>) 352000 98400 tensor(5.2040, device='cuda:0', grad_fn=<NllLossBackward0>) 98500 tensor(5.2696, device='cuda:0', grad_fn=<NllLossBackward0>) 353000 98600 tensor(5.3089, device='cuda:0', grad_fn=<NllLossBackward0>) 98700 tensor(5.0973, device='cuda:0', grad_fn=<NllLossBackward0>) 98800 tensor(5.0995, device='cuda:0', grad_fn=<NllLossBackward0>) 354000 98900 tensor(5.1192, device='cuda:0', grad_fn=<NllLossBackward0>) 99000 tensor(5.2984, device='cuda:0', grad_fn=<NllLossBackward0>) 99100 tensor(5.1856, device='cuda:0', grad_fn=<NllLossBackward0>) 355000 99200 tensor(5.1883, device='cuda:0', grad_fn=<NllLossBackward0>) 99300 tensor(4.9773, device='cuda:0', grad_fn=<NllLossBackward0>) 99400 tensor(5.2897, device='cuda:0', grad_fn=<NllLossBackward0>) 356000 99500 tensor(5.3124, device='cuda:0', grad_fn=<NllLossBackward0>) 99600 tensor(5.0386, device='cuda:0', grad_fn=<NllLossBackward0>) 99700 tensor(4.8906, device='cuda:0', grad_fn=<NllLossBackward0>) 357000 99800 tensor(5.1151, device='cuda:0', grad_fn=<NllLossBackward0>) 99900 tensor(5.0333, device='cuda:0', grad_fn=<NllLossBackward0>) 100000 tensor(4.9526, device='cuda:0', grad_fn=<NllLossBackward0>) 358000 100100 tensor(5.4919, device='cuda:0', grad_fn=<NllLossBackward0>) 100200 tensor(4.7812, device='cuda:0', grad_fn=<NllLossBackward0>) 359000 100300 tensor(5.0547, device='cuda:0', grad_fn=<NllLossBackward0>) 100400 tensor(5.0233, device='cuda:0', grad_fn=<NllLossBackward0>) 100500 tensor(4.8942, device='cuda:0', grad_fn=<NllLossBackward0>) 360000 100600 tensor(4.9477, device='cuda:0', grad_fn=<NllLossBackward0>) 100700 tensor(4.9277, device='cuda:0', grad_fn=<NllLossBackward0>) 100800 tensor(5.2499, device='cuda:0', grad_fn=<NllLossBackward0>) 361000 100900 tensor(5.2162, device='cuda:0', grad_fn=<NllLossBackward0>) 101000 tensor(5.0550, device='cuda:0', grad_fn=<NllLossBackward0>) 101100 tensor(5.0957, device='cuda:0', grad_fn=<NllLossBackward0>) 362000 101200 tensor(5.0617, device='cuda:0', grad_fn=<NllLossBackward0>) 101300 tensor(5.3101, device='cuda:0', grad_fn=<NllLossBackward0>) 363000 101400 tensor(4.8387, device='cuda:0', grad_fn=<NllLossBackward0>) 101500 tensor(5.0914, device='cuda:0', grad_fn=<NllLossBackward0>) 101600 tensor(5.3198, device='cuda:0', grad_fn=<NllLossBackward0>) 364000 101700 tensor(4.9333, device='cuda:0', grad_fn=<NllLossBackward0>) 101800 tensor(4.7367, device='cuda:0', grad_fn=<NllLossBackward0>) 101900 tensor(4.9608, device='cuda:0', grad_fn=<NllLossBackward0>) 365000 102000 tensor(5.2333, device='cuda:0', grad_fn=<NllLossBackward0>) 102100 tensor(5.1444, device='cuda:0', grad_fn=<NllLossBackward0>) 102200 tensor(5.4328, device='cuda:0', grad_fn=<NllLossBackward0>) 366000 102300 tensor(5.2760, device='cuda:0', grad_fn=<NllLossBackward0>) 102400 tensor(5.2229, device='cuda:0', grad_fn=<NllLossBackward0>) 102500 tensor(5.2043, device='cuda:0', grad_fn=<NllLossBackward0>) 367000 102600 tensor(5.1444, device='cuda:0', grad_fn=<NllLossBackward0>) 102700 tensor(5.2240, device='cuda:0', grad_fn=<NllLossBackward0>) 368000 102800 tensor(5.1234, device='cuda:0', grad_fn=<NllLossBackward0>) 102900 tensor(4.8963, device='cuda:0', grad_fn=<NllLossBackward0>) 103000 tensor(5.0684, device='cuda:0', grad_fn=<NllLossBackward0>) 369000 103100 tensor(5.3707, device='cuda:0', grad_fn=<NllLossBackward0>) 103200 tensor(4.9859, device='cuda:0', grad_fn=<NllLossBackward0>) 103300 tensor(5.0649, device='cuda:0', grad_fn=<NllLossBackward0>) 370000 103400 tensor(5.0871, device='cuda:0', grad_fn=<NllLossBackward0>) 103500 tensor(4.9837, device='cuda:0', grad_fn=<NllLossBackward0>) 103600 tensor(5.1120, device='cuda:0', grad_fn=<NllLossBackward0>) 371000 103700 tensor(5.2086, device='cuda:0', grad_fn=<NllLossBackward0>) 103800 tensor(5.0741, device='cuda:0', grad_fn=<NllLossBackward0>) 103900 tensor(4.9891, device='cuda:0', grad_fn=<NllLossBackward0>) 372000 104000 tensor(4.9637, device='cuda:0', grad_fn=<NllLossBackward0>) 104100 tensor(5.0033, device='cuda:0', grad_fn=<NllLossBackward0>) 373000 104200 tensor(5.1105, device='cuda:0', grad_fn=<NllLossBackward0>) 104300 tensor(4.9057, device='cuda:0', grad_fn=<NllLossBackward0>) 104400 tensor(5.2394, device='cuda:0', grad_fn=<NllLossBackward0>) 374000 104500 tensor(5.1488, device='cuda:0', grad_fn=<NllLossBackward0>) 104600 tensor(5.0034, device='cuda:0', grad_fn=<NllLossBackward0>) 104700 tensor(5.0169, device='cuda:0', grad_fn=<NllLossBackward0>) 375000 104800 tensor(5.5932, device='cuda:0', grad_fn=<NllLossBackward0>) 104900 tensor(5.0594, device='cuda:0', grad_fn=<NllLossBackward0>) 105000 tensor(5.1818, device='cuda:0', grad_fn=<NllLossBackward0>) 376000 105100 tensor(5.1265, device='cuda:0', grad_fn=<NllLossBackward0>) 105200 tensor(5.2194, device='cuda:0', grad_fn=<NllLossBackward0>) 105300 tensor(5.1688, device='cuda:0', grad_fn=<NllLossBackward0>) 377000 105400 tensor(5.3924, device='cuda:0', grad_fn=<NllLossBackward0>) 105500 tensor(5.2369, device='cuda:0', grad_fn=<NllLossBackward0>) 378000 105600 tensor(5.0005, device='cuda:0', grad_fn=<NllLossBackward0>) 105700 tensor(4.9490, device='cuda:0', grad_fn=<NllLossBackward0>) 105800 tensor(4.9947, device='cuda:0', grad_fn=<NllLossBackward0>) 379000 105900 tensor(5.1400, device='cuda:0', grad_fn=<NllLossBackward0>) 106000 tensor(5.1712, device='cuda:0', grad_fn=<NllLossBackward0>) 106100 tensor(5.3889, device='cuda:0', grad_fn=<NllLossBackward0>) 380000 106200 tensor(5.0640, device='cuda:0', grad_fn=<NllLossBackward0>) 106300 tensor(5.1624, device='cuda:0', grad_fn=<NllLossBackward0>) 106400 tensor(4.9938, device='cuda:0', grad_fn=<NllLossBackward0>) 381000 106500 tensor(5.2659, device='cuda:0', grad_fn=<NllLossBackward0>) 106600 tensor(4.9500, device='cuda:0', grad_fn=<NllLossBackward0>) 106700 tensor(5.2001, device='cuda:0', grad_fn=<NllLossBackward0>) 382000 106800 tensor(4.9776, device='cuda:0', grad_fn=<NllLossBackward0>) 106900 tensor(5.2684, device='cuda:0', grad_fn=<NllLossBackward0>) 383000 107000 tensor(4.9327, device='cuda:0', grad_fn=<NllLossBackward0>) 107100 tensor(5.1739, device='cuda:0', grad_fn=<NllLossBackward0>) 107200 tensor(5.1641, device='cuda:0', grad_fn=<NllLossBackward0>) 384000 107300 tensor(4.9207, device='cuda:0', grad_fn=<NllLossBackward0>) 107400 tensor(5.1112, device='cuda:0', grad_fn=<NllLossBackward0>) 107500 tensor(4.8839, device='cuda:0', grad_fn=<NllLossBackward0>) 385000 107600 tensor(4.9063, device='cuda:0', grad_fn=<NllLossBackward0>) 107700 tensor(5.1218, device='cuda:0', grad_fn=<NllLossBackward0>) 107800 tensor(5.1329, device='cuda:0', grad_fn=<NllLossBackward0>) 386000 107900 tensor(5.3598, device='cuda:0', grad_fn=<NllLossBackward0>) 108000 tensor(5.1047, device='cuda:0', grad_fn=<NllLossBackward0>) 108100 tensor(5.0709, device='cuda:0', grad_fn=<NllLossBackward0>) 387000 108200 tensor(5.0716, device='cuda:0', grad_fn=<NllLossBackward0>) 108300 tensor(5.0979, device='cuda:0', grad_fn=<NllLossBackward0>) 388000 108400 tensor(5.2054, device='cuda:0', grad_fn=<NllLossBackward0>) 108500 tensor(4.8856, device='cuda:0', grad_fn=<NllLossBackward0>) 108600 tensor(5.0763, device='cuda:0', grad_fn=<NllLossBackward0>) 389000 108700 tensor(4.6029, device='cuda:0', grad_fn=<NllLossBackward0>) 108800 tensor(4.9980, device='cuda:0', grad_fn=<NllLossBackward0>) 108900 tensor(4.7433, device='cuda:0', grad_fn=<NllLossBackward0>) 390000 109000 tensor(5.1747, device='cuda:0', grad_fn=<NllLossBackward0>) 109100 tensor(4.7590, device='cuda:0', grad_fn=<NllLossBackward0>) 109200 tensor(4.9555, device='cuda:0', grad_fn=<NllLossBackward0>) 391000 109300 tensor(4.6271, device='cuda:0', grad_fn=<NllLossBackward0>) 109400 tensor(4.9270, device='cuda:0', grad_fn=<NllLossBackward0>) 109500 tensor(4.8670, device='cuda:0', grad_fn=<NllLossBackward0>) 392000 109600 tensor(5.0673, device='cuda:0', grad_fn=<NllLossBackward0>) 109700 tensor(5.4325, device='cuda:0', grad_fn=<NllLossBackward0>) 393000 109800 tensor(4.7563, device='cuda:0', grad_fn=<NllLossBackward0>) 109900 tensor(4.9688, device='cuda:0', grad_fn=<NllLossBackward0>) 110000 tensor(5.0599, device='cuda:0', grad_fn=<NllLossBackward0>) 394000 110100 tensor(4.9771, device='cuda:0', grad_fn=<NllLossBackward0>) 110200 tensor(5.3440, device='cuda:0', grad_fn=<NllLossBackward0>) 110300 tensor(5.2067, device='cuda:0', grad_fn=<NllLossBackward0>) 395000 110400 tensor(4.6912, device='cuda:0', grad_fn=<NllLossBackward0>) 110500 tensor(5.1135, device='cuda:0', grad_fn=<NllLossBackward0>) 110600 tensor(4.9512, device='cuda:0', grad_fn=<NllLossBackward0>) 396000 110700 tensor(4.9049, device='cuda:0', grad_fn=<NllLossBackward0>) 110800 tensor(5.3380, device='cuda:0', grad_fn=<NllLossBackward0>) 110900 tensor(4.9982, device='cuda:0', grad_fn=<NllLossBackward0>) 397000 111000 tensor(5.1906, device='cuda:0', grad_fn=<NllLossBackward0>) 111100 tensor(4.9450, device='cuda:0', grad_fn=<NllLossBackward0>) 398000 111200 tensor(4.9843, device='cuda:0', grad_fn=<NllLossBackward0>) 111300 tensor(5.0615, device='cuda:0', grad_fn=<NllLossBackward0>) 111400 tensor(5.1630, device='cuda:0', grad_fn=<NllLossBackward0>) 399000 111500 tensor(5.2631, device='cuda:0', grad_fn=<NllLossBackward0>) 111600 tensor(5.2726, device='cuda:0', grad_fn=<NllLossBackward0>) 111700 tensor(4.8465, device='cuda:0', grad_fn=<NllLossBackward0>) 400000 111800 tensor(5.0638, device='cuda:0', grad_fn=<NllLossBackward0>) 111900 tensor(5.1322, device='cuda:0', grad_fn=<NllLossBackward0>) 112000 tensor(5.0956, device='cuda:0', grad_fn=<NllLossBackward0>) 401000 112100 tensor(4.6850, device='cuda:0', grad_fn=<NllLossBackward0>) 112200 tensor(5.0941, device='cuda:0', grad_fn=<NllLossBackward0>) 402000 112300 tensor(5.2451, device='cuda:0', grad_fn=<NllLossBackward0>) 112400 tensor(5.2617, device='cuda:0', grad_fn=<NllLossBackward0>) 112500 tensor(4.8904, device='cuda:0', grad_fn=<NllLossBackward0>) 403000 112600 tensor(4.9779, device='cuda:0', grad_fn=<NllLossBackward0>) 112700 tensor(4.9773, device='cuda:0', grad_fn=<NllLossBackward0>) 112800 tensor(4.6058, device='cuda:0', grad_fn=<NllLossBackward0>) 404000 112900 tensor(4.8841, device='cuda:0', grad_fn=<NllLossBackward0>) 113000 tensor(5.1203, device='cuda:0', grad_fn=<NllLossBackward0>) 113100 tensor(5.0650, device='cuda:0', grad_fn=<NllLossBackward0>) 405000 113200 tensor(4.6302, device='cuda:0', grad_fn=<NllLossBackward0>) 113300 tensor(4.9626, device='cuda:0', grad_fn=<NllLossBackward0>) 113400 tensor(5.0637, device='cuda:0', grad_fn=<NllLossBackward0>) 406000 113500 tensor(5.3193, device='cuda:0', grad_fn=<NllLossBackward0>) 113600 tensor(5.2732, device='cuda:0', grad_fn=<NllLossBackward0>) 407000 113700 tensor(4.8750, device='cuda:0', grad_fn=<NllLossBackward0>) 113800 tensor(5.2875, device='cuda:0', grad_fn=<NllLossBackward0>) 113900 tensor(5.0677, device='cuda:0', grad_fn=<NllLossBackward0>) 408000 114000 tensor(4.7401, device='cuda:0', grad_fn=<NllLossBackward0>) 114100 tensor(4.8493, device='cuda:0', grad_fn=<NllLossBackward0>) 114200 tensor(4.8203, device='cuda:0', grad_fn=<NllLossBackward0>) 409000 114300 tensor(4.6964, device='cuda:0', grad_fn=<NllLossBackward0>) 114400 tensor(4.9045, device='cuda:0', grad_fn=<NllLossBackward0>) 114500 tensor(4.9766, device='cuda:0', grad_fn=<NllLossBackward0>) 410000 114600 tensor(5.2010, device='cuda:0', grad_fn=<NllLossBackward0>) 114700 tensor(5.1521, device='cuda:0', grad_fn=<NllLossBackward0>) 114800 tensor(5.0945, device='cuda:0', grad_fn=<NllLossBackward0>) 411000 114900 tensor(5.0311, device='cuda:0', grad_fn=<NllLossBackward0>) 115000 tensor(5.0057, device='cuda:0', grad_fn=<NllLossBackward0>) 412000 115100 tensor(5.1129, device='cuda:0', grad_fn=<NllLossBackward0>) 115200 tensor(5.0684, device='cuda:0', grad_fn=<NllLossBackward0>) 115300 tensor(5.1183, device='cuda:0', grad_fn=<NllLossBackward0>) 413000 115400 tensor(5.3141, device='cuda:0', grad_fn=<NllLossBackward0>) 115500 tensor(5.3232, device='cuda:0', grad_fn=<NllLossBackward0>) 115600 tensor(5.1170, device='cuda:0', grad_fn=<NllLossBackward0>) 414000 115700 tensor(5.0023, device='cuda:0', grad_fn=<NllLossBackward0>) 115800 tensor(5.1229, device='cuda:0', grad_fn=<NllLossBackward0>) 115900 tensor(5.1923, device='cuda:0', grad_fn=<NllLossBackward0>) 415000 116000 tensor(5.1286, device='cuda:0', grad_fn=<NllLossBackward0>) 116100 tensor(4.8629, device='cuda:0', grad_fn=<NllLossBackward0>) 116200 tensor(5.2643, device='cuda:0', grad_fn=<NllLossBackward0>) 416000 116300 tensor(4.4037, device='cuda:0', grad_fn=<NllLossBackward0>) 116400 tensor(5.1548, device='cuda:0', grad_fn=<NllLossBackward0>) 417000 116500 tensor(5.0993, device='cuda:0', grad_fn=<NllLossBackward0>) 116600 tensor(5.1907, device='cuda:0', grad_fn=<NllLossBackward0>) 116700 tensor(5.0096, device='cuda:0', grad_fn=<NllLossBackward0>) 418000 116800 tensor(4.8711, device='cuda:0', grad_fn=<NllLossBackward0>) 116900 tensor(5.1060, device='cuda:0', grad_fn=<NllLossBackward0>) 117000 tensor(5.0866, device='cuda:0', grad_fn=<NllLossBackward0>) 419000 117100 tensor(5.1294, device='cuda:0', grad_fn=<NllLossBackward0>) 117200 tensor(4.9697, device='cuda:0', grad_fn=<NllLossBackward0>) 117300 tensor(5.3102, device='cuda:0', grad_fn=<NllLossBackward0>) 420000 117400 tensor(5.3132, device='cuda:0', grad_fn=<NllLossBackward0>) 117500 tensor(4.8965, device='cuda:0', grad_fn=<NllLossBackward0>) 421000 117600 tensor(5.2093, device='cuda:0', grad_fn=<NllLossBackward0>) 117700 tensor(5.1284, device='cuda:0', grad_fn=<NllLossBackward0>) 117800 tensor(4.9976, device='cuda:0', grad_fn=<NllLossBackward0>) 422000 117900 tensor(5.1066, device='cuda:0', grad_fn=<NllLossBackward0>) 118000 tensor(5.1549, device='cuda:0', grad_fn=<NllLossBackward0>) 118100 tensor(4.9796, device='cuda:0', grad_fn=<NllLossBackward0>) 423000 118200 tensor(5.0112, device='cuda:0', grad_fn=<NllLossBackward0>) 118300 tensor(5.0054, device='cuda:0', grad_fn=<NllLossBackward0>) 118400 tensor(5.1148, device='cuda:0', grad_fn=<NllLossBackward0>) 424000 118500 tensor(5.0216, device='cuda:0', grad_fn=<NllLossBackward0>) 118600 tensor(5.1426, device='cuda:0', grad_fn=<NllLossBackward0>) 118700 tensor(5.1632, device='cuda:0', grad_fn=<NllLossBackward0>) 425000 118800 tensor(5.1688, device='cuda:0', grad_fn=<NllLossBackward0>) 118900 tensor(5.2083, device='cuda:0', grad_fn=<NllLossBackward0>) 426000 119000 tensor(4.7053, device='cuda:0', grad_fn=<NllLossBackward0>) 119100 tensor(4.9990, device='cuda:0', grad_fn=<NllLossBackward0>) 119200 tensor(5.3487, device='cuda:0', grad_fn=<NllLossBackward0>) 427000 119300 tensor(4.9779, device='cuda:0', grad_fn=<NllLossBackward0>) 119400 tensor(5.0351, device='cuda:0', grad_fn=<NllLossBackward0>) 119500 tensor(5.0373, device='cuda:0', grad_fn=<NllLossBackward0>) 428000 119600 tensor(5.3674, device='cuda:0', grad_fn=<NllLossBackward0>) 119700 tensor(4.9741, device='cuda:0', grad_fn=<NllLossBackward0>) 119800 tensor(5.0577, device='cuda:0', grad_fn=<NllLossBackward0>) 429000 119900 tensor(4.8425, device='cuda:0', grad_fn=<NllLossBackward0>) 120000 tensor(4.9597, device='cuda:0', grad_fn=<NllLossBackward0>) 120100 tensor(4.9721, device='cuda:0', grad_fn=<NllLossBackward0>) 430000 120200 tensor(5.1214, device='cuda:0', grad_fn=<NllLossBackward0>) 120300 tensor(5.0225, device='cuda:0', grad_fn=<NllLossBackward0>) 431000 120400 tensor(5.3908, device='cuda:0', grad_fn=<NllLossBackward0>) 120500 tensor(4.9955, device='cuda:0', grad_fn=<NllLossBackward0>) 120600 tensor(5.2588, device='cuda:0', grad_fn=<NllLossBackward0>) 432000 epoch: = 1 0 tensor(5.2093, device='cuda:0', grad_fn=<NllLossBackward0>) 100 tensor(5.3814, device='cuda:0', grad_fn=<NllLossBackward0>) 200 tensor(5.1460, device='cuda:0', grad_fn=<NllLossBackward0>) 1000 300 tensor(5.3199, device='cuda:0', grad_fn=<NllLossBackward0>) 400 tensor(5.1499, device='cuda:0', grad_fn=<NllLossBackward0>) 500 tensor(5.0303, device='cuda:0', grad_fn=<NllLossBackward0>) 2000 600 tensor(5.0671, device='cuda:0', grad_fn=<NllLossBackward0>) 700 tensor(5.0000, device='cuda:0', grad_fn=<NllLossBackward0>) 800 tensor(4.9863, device='cuda:0', grad_fn=<NllLossBackward0>) 3000 900 tensor(5.0532, device='cuda:0', grad_fn=<NllLossBackward0>) 1000 tensor(5.0634, device='cuda:0', grad_fn=<NllLossBackward0>) 1100 tensor(4.8666, device='cuda:0', grad_fn=<NllLossBackward0>) 4000 1200 tensor(4.8362, device='cuda:0', grad_fn=<NllLossBackward0>) 1300 tensor(5.1104, device='cuda:0', grad_fn=<NllLossBackward0>) 1400 tensor(5.2784, device='cuda:0', grad_fn=<NllLossBackward0>) 5000 1500 tensor(5.3135, device='cuda:0', grad_fn=<NllLossBackward0>) 1600 tensor(5.1400, device='cuda:0', grad_fn=<NllLossBackward0>) 6000 1700 tensor(5.2427, device='cuda:0', grad_fn=<NllLossBackward0>) 1800 tensor(5.2361, device='cuda:0', grad_fn=<NllLossBackward0>) 1900 tensor(5.0860, device='cuda:0', grad_fn=<NllLossBackward0>) 7000 2000 tensor(4.7754, device='cuda:0', grad_fn=<NllLossBackward0>) 2100 tensor(5.2257, device='cuda:0', grad_fn=<NllLossBackward0>) 2200 tensor(4.9228, device='cuda:0', grad_fn=<NllLossBackward0>) 8000 2300 tensor(5.2087, device='cuda:0', grad_fn=<NllLossBackward0>) 2400 tensor(4.6009, device='cuda:0', grad_fn=<NllLossBackward0>) 2500 tensor(5.0485, device='cuda:0', grad_fn=<NllLossBackward0>) 9000 2600 tensor(4.8077, device='cuda:0', grad_fn=<NllLossBackward0>) 2700 tensor(5.1105, device='cuda:0', grad_fn=<NllLossBackward0>) 2800 tensor(4.6892, device='cuda:0', grad_fn=<NllLossBackward0>) 10000 2900 tensor(4.9577, device='cuda:0', grad_fn=<NllLossBackward0>) 3000 tensor(5.0162, device='cuda:0', grad_fn=<NllLossBackward0>) 11000 3100 tensor(4.9243, device='cuda:0', grad_fn=<NllLossBackward0>) 3200 tensor(5.0965, device='cuda:0', grad_fn=<NllLossBackward0>) 3300 tensor(4.8072, device='cuda:0', grad_fn=<NllLossBackward0>) 12000 3400 tensor(5.0026, device='cuda:0', grad_fn=<NllLossBackward0>) 3500 tensor(4.8778, device='cuda:0', grad_fn=<NllLossBackward0>) 3600 tensor(4.7804, device='cuda:0', grad_fn=<NllLossBackward0>) 13000 3700 tensor(5.0536, device='cuda:0', grad_fn=<NllLossBackward0>) 3800 tensor(5.0348, device='cuda:0', grad_fn=<NllLossBackward0>) 3900 tensor(5.0932, device='cuda:0', grad_fn=<NllLossBackward0>) 14000 4000 tensor(4.8247, device='cuda:0', grad_fn=<NllLossBackward0>) 4100 tensor(5.2288, device='cuda:0', grad_fn=<NllLossBackward0>) 4200 tensor(4.7091, device='cuda:0', grad_fn=<NllLossBackward0>) 15000 4300 tensor(5.1466, device='cuda:0', grad_fn=<NllLossBackward0>) 4400 tensor(4.8608, device='cuda:0', grad_fn=<NllLossBackward0>) 16000 4500 tensor(5.0831, device='cuda:0', grad_fn=<NllLossBackward0>) 4600 tensor(5.1534, device='cuda:0', grad_fn=<NllLossBackward0>) 4700 tensor(5.1433, device='cuda:0', grad_fn=<NllLossBackward0>) 17000 4800 tensor(5.0023, device='cuda:0', grad_fn=<NllLossBackward0>) 4900 tensor(5.1541, device='cuda:0', grad_fn=<NllLossBackward0>) 5000 tensor(5.2215, device='cuda:0', grad_fn=<NllLossBackward0>) 18000 5100 tensor(5.0103, device='cuda:0', grad_fn=<NllLossBackward0>) 5200 tensor(5.2232, device='cuda:0', grad_fn=<NllLossBackward0>) 5300 tensor(5.0542, device='cuda:0', grad_fn=<NllLossBackward0>) 19000 5400 tensor(4.9754, device='cuda:0', grad_fn=<NllLossBackward0>) 5500 tensor(5.1588, device='cuda:0', grad_fn=<NllLossBackward0>) 20000 5600 tensor(4.9750, device='cuda:0', grad_fn=<NllLossBackward0>) 5700 tensor(5.0814, device='cuda:0', grad_fn=<NllLossBackward0>) 5800 tensor(5.4650, device='cuda:0', grad_fn=<NllLossBackward0>) 21000 5900 tensor(5.2366, device='cuda:0', grad_fn=<NllLossBackward0>) 6000 tensor(5.2223, device='cuda:0', grad_fn=<NllLossBackward0>) 6100 tensor(4.9001, device='cuda:0', grad_fn=<NllLossBackward0>) 22000 6200 tensor(5.3510, device='cuda:0', grad_fn=<NllLossBackward0>) 6300 tensor(5.2366, device='cuda:0', grad_fn=<NllLossBackward0>) 6400 tensor(5.0379, device='cuda:0', grad_fn=<NllLossBackward0>) 23000 6500 tensor(5.0265, device='cuda:0', grad_fn=<NllLossBackward0>) 6600 tensor(5.5000, device='cuda:0', grad_fn=<NllLossBackward0>) 6700 tensor(4.8587, device='cuda:0', grad_fn=<NllLossBackward0>) 24000 6800 tensor(4.8712, device='cuda:0', grad_fn=<NllLossBackward0>) 6900 tensor(5.1405, device='cuda:0', grad_fn=<NllLossBackward0>) 25000 7000 tensor(5.0429, device='cuda:0', grad_fn=<NllLossBackward0>) 7100 tensor(5.1420, device='cuda:0', grad_fn=<NllLossBackward0>) 7200 tensor(5.2794, device='cuda:0', grad_fn=<NllLossBackward0>) 26000 7300 tensor(5.2271, device='cuda:0', grad_fn=<NllLossBackward0>) 7400 tensor(4.9754, device='cuda:0', grad_fn=<NllLossBackward0>) 7500 tensor(5.3659, device='cuda:0', grad_fn=<NllLossBackward0>) 27000 7600 tensor(4.7902, device='cuda:0', grad_fn=<NllLossBackward0>) 7700 tensor(5.2327, device='cuda:0', grad_fn=<NllLossBackward0>) 7800 tensor(5.5528, device='cuda:0', grad_fn=<NllLossBackward0>) 28000 7900 tensor(5.0152, device='cuda:0', grad_fn=<NllLossBackward0>) 8000 tensor(5.1026, device='cuda:0', grad_fn=<NllLossBackward0>) 8100 tensor(4.6152, device='cuda:0', grad_fn=<NllLossBackward0>) 29000 8200 tensor(5.0953, device='cuda:0', grad_fn=<NllLossBackward0>) 8300 tensor(5.0372, device='cuda:0', grad_fn=<NllLossBackward0>) 30000 8400 tensor(5.0211, device='cuda:0', grad_fn=<NllLossBackward0>) 8500 tensor(4.7475, device='cuda:0', grad_fn=<NllLossBackward0>) 8600 tensor(4.8443, device='cuda:0', grad_fn=<NllLossBackward0>) 31000 8700 tensor(5.1471, device='cuda:0', grad_fn=<NllLossBackward0>) 8800 tensor(5.1762, device='cuda:0', grad_fn=<NllLossBackward0>) 8900 tensor(5.4800, device='cuda:0', grad_fn=<NllLossBackward0>) 32000 9000 tensor(4.9010, device='cuda:0', grad_fn=<NllLossBackward0>) 9100 tensor(5.1490, device='cuda:0', grad_fn=<NllLossBackward0>) 9200 tensor(5.2119, device='cuda:0', grad_fn=<NllLossBackward0>) 33000 9300 tensor(5.0652, device='cuda:0', grad_fn=<NllLossBackward0>) 9400 tensor(5.0831, device='cuda:0', grad_fn=<NllLossBackward0>) 34000 9500 tensor(4.9114, device='cuda:0', grad_fn=<NllLossBackward0>) 9600 tensor(4.9139, device='cuda:0', grad_fn=<NllLossBackward0>) 9700 tensor(5.0307, device='cuda:0', grad_fn=<NllLossBackward0>) 35000 9800 tensor(5.1592, device='cuda:0', grad_fn=<NllLossBackward0>) 9900 tensor(4.9091, device='cuda:0', grad_fn=<NllLossBackward0>) 10000 tensor(5.2556, device='cuda:0', grad_fn=<NllLossBackward0>) 36000 10100 tensor(5.2265, device='cuda:0', grad_fn=<NllLossBackward0>) 10200 tensor(5.0764, device='cuda:0', grad_fn=<NllLossBackward0>) 10300 tensor(5.2197, device='cuda:0', grad_fn=<NllLossBackward0>) 37000 10400 tensor(4.9872, device='cuda:0', grad_fn=<NllLossBackward0>) 10500 tensor(5.0030, device='cuda:0', grad_fn=<NllLossBackward0>) 10600 tensor(4.7093, device='cuda:0', grad_fn=<NllLossBackward0>) 38000 10700 tensor(5.1943, device='cuda:0', grad_fn=<NllLossBackward0>) 10800 tensor(4.6028, device='cuda:0', grad_fn=<NllLossBackward0>) 39000 10900 tensor(5.1688, device='cuda:0', grad_fn=<NllLossBackward0>) 11000 tensor(4.9071, device='cuda:0', grad_fn=<NllLossBackward0>) 11100 tensor(4.7852, device='cuda:0', grad_fn=<NllLossBackward0>) 40000 11200 tensor(5.1724, device='cuda:0', grad_fn=<NllLossBackward0>) 11300 tensor(5.1048, device='cuda:0', grad_fn=<NllLossBackward0>) 11400 tensor(5.3444, device='cuda:0', grad_fn=<NllLossBackward0>) 41000 11500 tensor(5.2196, device='cuda:0', grad_fn=<NllLossBackward0>) 11600 tensor(4.8939, device='cuda:0', grad_fn=<NllLossBackward0>) 11700 tensor(5.2069, device='cuda:0', grad_fn=<NllLossBackward0>) 42000 11800 tensor(5.3687, device='cuda:0', grad_fn=<NllLossBackward0>) 11900 tensor(4.9033, device='cuda:0', grad_fn=<NllLossBackward0>) 12000 tensor(5.1747, device='cuda:0', grad_fn=<NllLossBackward0>) 43000 12100 tensor(5.0138, device='cuda:0', grad_fn=<NllLossBackward0>) 12200 tensor(5.0538, device='cuda:0', grad_fn=<NllLossBackward0>) 44000 12300 tensor(5.2248, device='cuda:0', grad_fn=<NllLossBackward0>) 12400 tensor(4.8343, device='cuda:0', grad_fn=<NllLossBackward0>) 12500 tensor(5.0820, device='cuda:0', grad_fn=<NllLossBackward0>) 45000 12600 tensor(5.0378, device='cuda:0', grad_fn=<NllLossBackward0>) 12700 tensor(4.7855, device='cuda:0', grad_fn=<NllLossBackward0>) 12800 tensor(5.1147, device='cuda:0', grad_fn=<NllLossBackward0>) 46000 12900 tensor(5.2598, device='cuda:0', grad_fn=<NllLossBackward0>) 13000 tensor(4.8743, device='cuda:0', grad_fn=<NllLossBackward0>) 13100 tensor(5.1137, device='cuda:0', grad_fn=<NllLossBackward0>) 47000 13200 tensor(5.2270, device='cuda:0', grad_fn=<NllLossBackward0>) 13300 tensor(5.3398, device='cuda:0', grad_fn=<NllLossBackward0>) 13400 tensor(4.7988, device='cuda:0', grad_fn=<NllLossBackward0>) 48000 13500 tensor(5.1067, device='cuda:0', grad_fn=<NllLossBackward0>) 13600 tensor(5.0730, device='cuda:0', grad_fn=<NllLossBackward0>) 49000 13700 tensor(5.1172, device='cuda:0', grad_fn=<NllLossBackward0>) 13800 tensor(4.9974, device='cuda:0', grad_fn=<NllLossBackward0>) 13900 tensor(5.0809, device='cuda:0', grad_fn=<NllLossBackward0>) 50000 14000 tensor(5.0515, device='cuda:0', grad_fn=<NllLossBackward0>) 14100 tensor(5.2082, device='cuda:0', grad_fn=<NllLossBackward0>) 14200 tensor(5.2196, device='cuda:0', grad_fn=<NllLossBackward0>) 51000 14300 tensor(5.1963, device='cuda:0', grad_fn=<NllLossBackward0>) 14400 tensor(4.6420, device='cuda:0', grad_fn=<NllLossBackward0>) 14500 tensor(4.7768, device='cuda:0', grad_fn=<NllLossBackward0>) 52000 14600 tensor(4.8853, device='cuda:0', grad_fn=<NllLossBackward0>) 14700 tensor(5.2161, device='cuda:0', grad_fn=<NllLossBackward0>) 14800 tensor(4.8348, device='cuda:0', grad_fn=<NllLossBackward0>) 53000 14900 tensor(4.7723, device='cuda:0', grad_fn=<NllLossBackward0>) 15000 tensor(5.1951, device='cuda:0', grad_fn=<NllLossBackward0>) 54000 15100 tensor(4.5943, device='cuda:0', grad_fn=<NllLossBackward0>) 15200 tensor(4.8785, device='cuda:0', grad_fn=<NllLossBackward0>) 15300 tensor(5.2401, device='cuda:0', grad_fn=<NllLossBackward0>) 55000 15400 tensor(5.1345, device='cuda:0', grad_fn=<NllLossBackward0>) 15500 tensor(4.9845, device='cuda:0', grad_fn=<NllLossBackward0>) 15600 tensor(5.0955, device='cuda:0', grad_fn=<NllLossBackward0>) 56000 15700 tensor(5.0750, device='cuda:0', grad_fn=<NllLossBackward0>) 15800 tensor(4.8953, device='cuda:0', grad_fn=<NllLossBackward0>) 15900 tensor(5.0173, device='cuda:0', grad_fn=<NllLossBackward0>) 57000 16000 tensor(5.3809, device='cuda:0', grad_fn=<NllLossBackward0>) 16100 tensor(5.2108, device='cuda:0', grad_fn=<NllLossBackward0>) 16200 tensor(4.9123, device='cuda:0', grad_fn=<NllLossBackward0>) 58000 16300 tensor(4.8491, device='cuda:0', grad_fn=<NllLossBackward0>) 16400 tensor(5.1661, device='cuda:0', grad_fn=<NllLossBackward0>) 59000 16500 tensor(5.0995, device='cuda:0', grad_fn=<NllLossBackward0>) 16600 tensor(5.1714, device='cuda:0', grad_fn=<NllLossBackward0>) 16700 tensor(5.0571, device='cuda:0', grad_fn=<NllLossBackward0>) 60000 16800 tensor(5.1657, device='cuda:0', grad_fn=<NllLossBackward0>) 16900 tensor(5.1391, device='cuda:0', grad_fn=<NllLossBackward0>) 17000 tensor(5.1067, device='cuda:0', grad_fn=<NllLossBackward0>) 61000 17100 tensor(5.1680, device='cuda:0', grad_fn=<NllLossBackward0>) 17200 tensor(5.2227, device='cuda:0', grad_fn=<NllLossBackward0>) 17300 tensor(4.6891, device='cuda:0', grad_fn=<NllLossBackward0>) 62000 17400 tensor(5.0639, device='cuda:0', grad_fn=<NllLossBackward0>) 17500 tensor(5.0507, device='cuda:0', grad_fn=<NllLossBackward0>) 17600 tensor(4.9053, device='cuda:0', grad_fn=<NllLossBackward0>) 63000 17700 tensor(4.8750, device='cuda:0', grad_fn=<NllLossBackward0>) 17800 tensor(4.7771, device='cuda:0', grad_fn=<NllLossBackward0>) 64000 17900 tensor(4.9931, device='cuda:0', grad_fn=<NllLossBackward0>) 18000 tensor(5.1721, device='cuda:0', grad_fn=<NllLossBackward0>) 18100 tensor(5.0419, device='cuda:0', grad_fn=<NllLossBackward0>) 65000 18200 tensor(5.0401, device='cuda:0', grad_fn=<NllLossBackward0>) 18300 tensor(4.9477, device='cuda:0', grad_fn=<NllLossBackward0>) 18400 tensor(4.9341, device='cuda:0', grad_fn=<NllLossBackward0>) 66000 18500 tensor(5.0645, device='cuda:0', grad_fn=<NllLossBackward0>) 18600 tensor(5.1702, device='cuda:0', grad_fn=<NllLossBackward0>) 18700 tensor(5.0680, device='cuda:0', grad_fn=<NllLossBackward0>) 67000 18800 tensor(4.7324, device='cuda:0', grad_fn=<NllLossBackward0>) 18900 tensor(5.1745, device='cuda:0', grad_fn=<NllLossBackward0>) 68000 19000 tensor(5.0148, device='cuda:0', grad_fn=<NllLossBackward0>) 19100 tensor(4.9552, device='cuda:0', grad_fn=<NllLossBackward0>) 19200 tensor(5.0249, device='cuda:0', grad_fn=<NllLossBackward0>) 69000 19300 tensor(5.0709, device='cuda:0', grad_fn=<NllLossBackward0>) 19400 tensor(5.1439, device='cuda:0', grad_fn=<NllLossBackward0>) 19500 tensor(4.9734, device='cuda:0', grad_fn=<NllLossBackward0>) 70000 19600 tensor(5.2309, device='cuda:0', grad_fn=<NllLossBackward0>) 19700 tensor(5.0292, device='cuda:0', grad_fn=<NllLossBackward0>) 19800 tensor(4.4443, device='cuda:0', grad_fn=<NllLossBackward0>) 71000 19900 tensor(5.2148, device='cuda:0', grad_fn=<NllLossBackward0>) 20000 tensor(4.8955, device='cuda:0', grad_fn=<NllLossBackward0>) 20100 tensor(4.9404, device='cuda:0', grad_fn=<NllLossBackward0>) 72000 20200 tensor(5.0525, device='cuda:0', grad_fn=<NllLossBackward0>) 20300 tensor(5.2610, device='cuda:0', grad_fn=<NllLossBackward0>) 73000 20400 tensor(5.2233, device='cuda:0', grad_fn=<NllLossBackward0>) 20500 tensor(5.1364, device='cuda:0', grad_fn=<NllLossBackward0>) 20600 tensor(4.9917, device='cuda:0', grad_fn=<NllLossBackward0>) 74000 20700 tensor(4.9608, device='cuda:0', grad_fn=<NllLossBackward0>) 20800 tensor(5.1992, device='cuda:0', grad_fn=<NllLossBackward0>) 20900 tensor(5.1686, device='cuda:0', grad_fn=<NllLossBackward0>) 75000 21000 tensor(4.8483, device='cuda:0', grad_fn=<NllLossBackward0>) 21100 tensor(5.4130, device='cuda:0', grad_fn=<NllLossBackward0>) 21200 tensor(4.7093, device='cuda:0', grad_fn=<NllLossBackward0>) 76000 21300 tensor(5.2148, device='cuda:0', grad_fn=<NllLossBackward0>) 21400 tensor(4.9624, device='cuda:0', grad_fn=<NllLossBackward0>) 21500 tensor(4.9292, device='cuda:0', grad_fn=<NllLossBackward0>) 77000 21600 tensor(5.1661, device='cuda:0', grad_fn=<NllLossBackward0>) 21700 tensor(5.0725, device='cuda:0', grad_fn=<NllLossBackward0>) 21800 tensor(4.7123, device='cuda:0', grad_fn=<NllLossBackward0>) 78000 21900 tensor(5.2135, device='cuda:0', grad_fn=<NllLossBackward0>) 22000 tensor(4.8017, device='cuda:0', grad_fn=<NllLossBackward0>) 79000 22100 tensor(5.1210, device='cuda:0', grad_fn=<NllLossBackward0>) 22200 tensor(5.0554, device='cuda:0', grad_fn=<NllLossBackward0>) 22300 tensor(5.0997, device='cuda:0', grad_fn=<NllLossBackward0>) 80000 22400 tensor(4.8383, device='cuda:0', grad_fn=<NllLossBackward0>) 22500 tensor(4.5534, device='cuda:0', grad_fn=<NllLossBackward0>) 22600 tensor(5.1061, device='cuda:0', grad_fn=<NllLossBackward0>) 81000 22700 tensor(4.9188, device='cuda:0', grad_fn=<NllLossBackward0>) 22800 tensor(5.0239, device='cuda:0', grad_fn=<NllLossBackward0>) 22900 tensor(4.9186, device='cuda:0', grad_fn=<NllLossBackward0>) 82000 23000 tensor(5.3097, device='cuda:0', grad_fn=<NllLossBackward0>) 23100 tensor(5.0750, device='cuda:0', grad_fn=<NllLossBackward0>) 83000 23200 tensor(4.9890, device='cuda:0', grad_fn=<NllLossBackward0>) 23300 tensor(5.1193, device='cuda:0', grad_fn=<NllLossBackward0>) 23400 tensor(5.0708, device='cuda:0', grad_fn=<NllLossBackward0>) 84000 23500 tensor(4.9966, device='cuda:0', grad_fn=<NllLossBackward0>) 23600 tensor(5.1132, device='cuda:0', grad_fn=<NllLossBackward0>) 23700 tensor(5.3187, device='cuda:0', grad_fn=<NllLossBackward0>) 85000 23800 tensor(5.2937, device='cuda:0', grad_fn=<NllLossBackward0>) 23900 tensor(5.2049, device='cuda:0', grad_fn=<NllLossBackward0>) 24000 tensor(5.0551, device='cuda:0', grad_fn=<NllLossBackward0>) 86000 24100 tensor(5.3820, device='cuda:0', grad_fn=<NllLossBackward0>) 24200 tensor(5.2243, device='cuda:0', grad_fn=<NllLossBackward0>) 24300 tensor(5.0407, device='cuda:0', grad_fn=<NllLossBackward0>) 87000 24400 tensor(5.1378, device='cuda:0', grad_fn=<NllLossBackward0>) 24500 tensor(4.8982, device='cuda:0', grad_fn=<NllLossBackward0>) 88000 24600 tensor(5.0213, device='cuda:0', grad_fn=<NllLossBackward0>) 24700 tensor(5.1981, device='cuda:0', grad_fn=<NllLossBackward0>) 24800 tensor(5.2277, device='cuda:0', grad_fn=<NllLossBackward0>) 89000 24900 tensor(5.1798, device='cuda:0', grad_fn=<NllLossBackward0>) 25000 tensor(5.4307, device='cuda:0', grad_fn=<NllLossBackward0>) 25100 tensor(5.0697, device='cuda:0', grad_fn=<NllLossBackward0>) 90000 25200 tensor(5.1689, device='cuda:0', grad_fn=<NllLossBackward0>) 25300 tensor(5.1092, device='cuda:0', grad_fn=<NllLossBackward0>) 25400 tensor(5.0354, device='cuda:0', grad_fn=<NllLossBackward0>) 91000 25500 tensor(4.9247, device='cuda:0', grad_fn=<NllLossBackward0>) 25600 tensor(4.6927, device='cuda:0', grad_fn=<NllLossBackward0>) 25700 tensor(5.0795, device='cuda:0', grad_fn=<NllLossBackward0>) 92000 25800 tensor(5.0327, device='cuda:0', grad_fn=<NllLossBackward0>) 25900 tensor(5.1922, device='cuda:0', grad_fn=<NllLossBackward0>) 93000 26000 tensor(4.7922, device='cuda:0', grad_fn=<NllLossBackward0>) 26100 tensor(5.2135, device='cuda:0', grad_fn=<NllLossBackward0>) 26200 tensor(4.8716, device='cuda:0', grad_fn=<NllLossBackward0>) 94000 26300 tensor(5.1197, device='cuda:0', grad_fn=<NllLossBackward0>) 26400 tensor(5.0293, device='cuda:0', grad_fn=<NllLossBackward0>) 26500 tensor(5.0354, device='cuda:0', grad_fn=<NllLossBackward0>) 95000 26600 tensor(5.2529, device='cuda:0', grad_fn=<NllLossBackward0>) 26700 tensor(5.0883, device='cuda:0', grad_fn=<NllLossBackward0>) 26800 tensor(4.8198, device='cuda:0', grad_fn=<NllLossBackward0>) 96000 26900 tensor(5.0282, device='cuda:0', grad_fn=<NllLossBackward0>) 27000 tensor(4.7931, device='cuda:0', grad_fn=<NllLossBackward0>) 27100 tensor(5.0120, device='cuda:0', grad_fn=<NllLossBackward0>) 97000 27200 tensor(5.0787, device='cuda:0', grad_fn=<NllLossBackward0>) 27300 tensor(5.0908, device='cuda:0', grad_fn=<NllLossBackward0>) 98000 27400 tensor(4.9436, device='cuda:0', grad_fn=<NllLossBackward0>) 27500 tensor(5.1993, device='cuda:0', grad_fn=<NllLossBackward0>) 27600 tensor(5.1542, device='cuda:0', grad_fn=<NllLossBackward0>) 99000 27700 tensor(5.1619, device='cuda:0', grad_fn=<NllLossBackward0>) 27800 tensor(4.9495, device='cuda:0', grad_fn=<NllLossBackward0>) 27900 tensor(4.9496, device='cuda:0', grad_fn=<NllLossBackward0>) 100000 28000 tensor(5.3698, device='cuda:0', grad_fn=<NllLossBackward0>) 28100 tensor(5.3364, device='cuda:0', grad_fn=<NllLossBackward0>) 28200 tensor(4.9353, device='cuda:0', grad_fn=<NllLossBackward0>) 101000 28300 tensor(5.1149, device='cuda:0', grad_fn=<NllLossBackward0>) 28400 tensor(4.7524, device='cuda:0', grad_fn=<NllLossBackward0>) 28500 tensor(5.0890, device='cuda:0', grad_fn=<NllLossBackward0>) 102000 28600 tensor(5.5074, device='cuda:0', grad_fn=<NllLossBackward0>) 28700 tensor(5.1043, device='cuda:0', grad_fn=<NllLossBackward0>) 103000 28800 tensor(5.0942, device='cuda:0', grad_fn=<NllLossBackward0>) 28900 tensor(4.7643, device='cuda:0', grad_fn=<NllLossBackward0>) 29000 tensor(4.9018, device='cuda:0', grad_fn=<NllLossBackward0>) 104000 29100 tensor(5.1374, device='cuda:0', grad_fn=<NllLossBackward0>) 29200 tensor(4.8248, device='cuda:0', grad_fn=<NllLossBackward0>) 29300 tensor(5.1806, device='cuda:0', grad_fn=<NllLossBackward0>) 105000 29400 tensor(5.1401, device='cuda:0', grad_fn=<NllLossBackward0>) 29500 tensor(4.9909, device='cuda:0', grad_fn=<NllLossBackward0>) 29600 tensor(5.3300, device='cuda:0', grad_fn=<NllLossBackward0>) 106000 29700 tensor(4.7358, device='cuda:0', grad_fn=<NllLossBackward0>) 29800 tensor(4.8462, device='cuda:0', grad_fn=<NllLossBackward0>) 29900 tensor(4.9805, device='cuda:0', grad_fn=<NllLossBackward0>) 107000 30000 tensor(5.1522, device='cuda:0', grad_fn=<NllLossBackward0>) 30100 tensor(5.0940, device='cuda:0', grad_fn=<NllLossBackward0>) 108000 30200 tensor(5.1651, device='cuda:0', grad_fn=<NllLossBackward0>) 30300 tensor(4.8124, device='cuda:0', grad_fn=<NllLossBackward0>) 30400 tensor(4.7609, device='cuda:0', grad_fn=<NllLossBackward0>) 109000 30500 tensor(4.8187, device='cuda:0', grad_fn=<NllLossBackward0>) 30600 tensor(4.9865, device='cuda:0', grad_fn=<NllLossBackward0>) 30700 tensor(5.0208, device='cuda:0', grad_fn=<NllLossBackward0>) 110000 30800 tensor(5.0904, device='cuda:0', grad_fn=<NllLossBackward0>) 30900 tensor(4.9448, device='cuda:0', grad_fn=<NllLossBackward0>) 31000 tensor(5.1119, device='cuda:0', grad_fn=<NllLossBackward0>) 111000 31100 tensor(5.1994, device='cuda:0', grad_fn=<NllLossBackward0>) 31200 tensor(5.0280, device='cuda:0', grad_fn=<NllLossBackward0>) 31300 tensor(5.2946, device='cuda:0', grad_fn=<NllLossBackward0>) 112000 31400 tensor(4.9557, device='cuda:0', grad_fn=<NllLossBackward0>) 31500 tensor(4.9296, device='cuda:0', grad_fn=<NllLossBackward0>) 113000 31600 tensor(4.8751, device='cuda:0', grad_fn=<NllLossBackward0>) 31700 tensor(5.3086, device='cuda:0', grad_fn=<NllLossBackward0>) 31800 tensor(4.7567, device='cuda:0', grad_fn=<NllLossBackward0>) 114000 31900 tensor(4.9941, device='cuda:0', grad_fn=<NllLossBackward0>) 32000 tensor(5.2035, device='cuda:0', grad_fn=<NllLossBackward0>) 32100 tensor(4.8145, device='cuda:0', grad_fn=<NllLossBackward0>) 115000 32200 tensor(5.3565, device='cuda:0', grad_fn=<NllLossBackward0>) 32300 tensor(4.9674, device='cuda:0', grad_fn=<NllLossBackward0>) 32400 tensor(4.9422, device='cuda:0', grad_fn=<NllLossBackward0>) 116000 32500 tensor(5.2489, device='cuda:0', grad_fn=<NllLossBackward0>) 32600 tensor(5.0207, device='cuda:0', grad_fn=<NllLossBackward0>) 117000 32700 tensor(5.1569, device='cuda:0', grad_fn=<NllLossBackward0>) 32800 tensor(5.0727, device='cuda:0', grad_fn=<NllLossBackward0>) 32900 tensor(4.9706, device='cuda:0', grad_fn=<NllLossBackward0>) 118000 33000 tensor(5.0288, device='cuda:0', grad_fn=<NllLossBackward0>) 33100 tensor(5.1344, device='cuda:0', grad_fn=<NllLossBackward0>) 33200 tensor(5.0739, device='cuda:0', grad_fn=<NllLossBackward0>) 119000 33300 tensor(4.7502, device='cuda:0', grad_fn=<NllLossBackward0>) 33400 tensor(4.6710, device='cuda:0', grad_fn=<NllLossBackward0>) 33500 tensor(5.3873, device='cuda:0', grad_fn=<NllLossBackward0>) 120000 33600 tensor(4.9222, device='cuda:0', grad_fn=<NllLossBackward0>) 33700 tensor(5.1264, device='cuda:0', grad_fn=<NllLossBackward0>) 33800 tensor(4.9766, device='cuda:0', grad_fn=<NllLossBackward0>) 121000 33900 tensor(5.0954, device='cuda:0', grad_fn=<NllLossBackward0>) 34000 tensor(5.0210, device='cuda:0', grad_fn=<NllLossBackward0>) 122000 34100 tensor(5.0876, device='cuda:0', grad_fn=<NllLossBackward0>) 34200 tensor(5.1302, device='cuda:0', grad_fn=<NllLossBackward0>) 34300 tensor(5.2983, device='cuda:0', grad_fn=<NllLossBackward0>) 123000 34400 tensor(5.2121, device='cuda:0', grad_fn=<NllLossBackward0>) 34500 tensor(5.3244, device='cuda:0', grad_fn=<NllLossBackward0>) 34600 tensor(4.9485, device='cuda:0', grad_fn=<NllLossBackward0>) 124000 34700 tensor(5.2090, device='cuda:0', grad_fn=<NllLossBackward0>) 34800 tensor(4.9158, device='cuda:0', grad_fn=<NllLossBackward0>) 34900 tensor(5.2443, device='cuda:0', grad_fn=<NllLossBackward0>) 125000 35000 tensor(5.1179, device='cuda:0', grad_fn=<NllLossBackward0>) 35100 tensor(5.0717, device='cuda:0', grad_fn=<NllLossBackward0>) 35200 tensor(5.1812, device='cuda:0', grad_fn=<NllLossBackward0>) 126000 35300 tensor(4.9081, device='cuda:0', grad_fn=<NllLossBackward0>) 35400 tensor(5.1653, device='cuda:0', grad_fn=<NllLossBackward0>) 127000 35500 tensor(5.1505, device='cuda:0', grad_fn=<NllLossBackward0>) 35600 tensor(5.2750, device='cuda:0', grad_fn=<NllLossBackward0>) 35700 tensor(5.4164, device='cuda:0', grad_fn=<NllLossBackward0>) 128000 35800 tensor(5.1043, device='cuda:0', grad_fn=<NllLossBackward0>) 35900 tensor(5.1016, device='cuda:0', grad_fn=<NllLossBackward0>) 36000 tensor(5.1461, device='cuda:0', grad_fn=<NllLossBackward0>) 129000 36100 tensor(5.2056, device='cuda:0', grad_fn=<NllLossBackward0>) 36200 tensor(5.1355, device='cuda:0', grad_fn=<NllLossBackward0>) 36300 tensor(5.3760, device='cuda:0', grad_fn=<NllLossBackward0>) 130000 36400 tensor(5.2054, device='cuda:0', grad_fn=<NllLossBackward0>) 36500 tensor(5.3691, device='cuda:0', grad_fn=<NllLossBackward0>) 131000 36600 tensor(4.8645, device='cuda:0', grad_fn=<NllLossBackward0>) 36700 tensor(5.4200, device='cuda:0', grad_fn=<NllLossBackward0>) 36800 tensor(5.2951, device='cuda:0', grad_fn=<NllLossBackward0>) 132000 36900 tensor(4.9001, device='cuda:0', grad_fn=<NllLossBackward0>) 37000 tensor(5.1626, device='cuda:0', grad_fn=<NllLossBackward0>) 37100 tensor(4.8412, device='cuda:0', grad_fn=<NllLossBackward0>) 133000 37200 tensor(4.7379, device='cuda:0', grad_fn=<NllLossBackward0>) 37300 tensor(4.8720, device='cuda:0', grad_fn=<NllLossBackward0>) 37400 tensor(4.8967, device='cuda:0', grad_fn=<NllLossBackward0>) 134000 37500 tensor(4.7772, device='cuda:0', grad_fn=<NllLossBackward0>) 37600 tensor(4.8302, device='cuda:0', grad_fn=<NllLossBackward0>) 37700 tensor(5.1201, device='cuda:0', grad_fn=<NllLossBackward0>) 135000 37800 tensor(5.1669, device='cuda:0', grad_fn=<NllLossBackward0>) 37900 tensor(5.2367, device='cuda:0', grad_fn=<NllLossBackward0>) 136000 38000 tensor(4.9588, device='cuda:0', grad_fn=<NllLossBackward0>) 38100 tensor(5.1460, device='cuda:0', grad_fn=<NllLossBackward0>) 38200 tensor(4.8495, device='cuda:0', grad_fn=<NllLossBackward0>) 137000 38300 tensor(5.2402, device='cuda:0', grad_fn=<NllLossBackward0>) 38400 tensor(4.9173, device='cuda:0', grad_fn=<NllLossBackward0>) 38500 tensor(5.1901, device='cuda:0', grad_fn=<NllLossBackward0>) 138000 38600 tensor(4.7848, device='cuda:0', grad_fn=<NllLossBackward0>) 38700 tensor(5.0920, device='cuda:0', grad_fn=<NllLossBackward0>) 38800 tensor(4.8625, device='cuda:0', grad_fn=<NllLossBackward0>) 139000 38900 tensor(4.9405, device='cuda:0', grad_fn=<NllLossBackward0>) 39000 tensor(5.1992, device='cuda:0', grad_fn=<NllLossBackward0>) 39100 tensor(4.8887, device='cuda:0', grad_fn=<NllLossBackward0>) 140000 39200 tensor(5.3122, device='cuda:0', grad_fn=<NllLossBackward0>) 39300 tensor(5.0380, device='cuda:0', grad_fn=<NllLossBackward0>) 141000 39400 tensor(4.9989, device='cuda:0', grad_fn=<NllLossBackward0>) 39500 tensor(5.1872, device='cuda:0', grad_fn=<NllLossBackward0>) 39600 tensor(4.9061, device='cuda:0', grad_fn=<NllLossBackward0>) 142000 39700 tensor(5.2725, device='cuda:0', grad_fn=<NllLossBackward0>) 39800 tensor(5.1865, device='cuda:0', grad_fn=<NllLossBackward0>) 39900 tensor(4.8060, device='cuda:0', grad_fn=<NllLossBackward0>) 143000 40000 tensor(4.9448, device='cuda:0', grad_fn=<NllLossBackward0>) 40100 tensor(5.0130, device='cuda:0', grad_fn=<NllLossBackward0>) 40200 tensor(5.0753, device='cuda:0', grad_fn=<NllLossBackward0>) 144000 40300 tensor(5.0743, device='cuda:0', grad_fn=<NllLossBackward0>) 40400 tensor(5.3216, device='cuda:0', grad_fn=<NllLossBackward0>) 40500 tensor(4.7829, device='cuda:0', grad_fn=<NllLossBackward0>) 145000 40600 tensor(5.3467, device='cuda:0', grad_fn=<NllLossBackward0>) 40700 tensor(5.2229, device='cuda:0', grad_fn=<NllLossBackward0>) 146000 40800 tensor(4.9148, device='cuda:0', grad_fn=<NllLossBackward0>) 40900 tensor(4.9961, device='cuda:0', grad_fn=<NllLossBackward0>) 41000 tensor(5.1233, device='cuda:0', grad_fn=<NllLossBackward0>) 147000 41100 tensor(4.9480, device='cuda:0', grad_fn=<NllLossBackward0>) 41200 tensor(5.1455, device='cuda:0', grad_fn=<NllLossBackward0>) 41300 tensor(4.7624, device='cuda:0', grad_fn=<NllLossBackward0>) 148000 41400 tensor(4.8317, device='cuda:0', grad_fn=<NllLossBackward0>) 41500 tensor(5.3050, device='cuda:0', grad_fn=<NllLossBackward0>) 41600 tensor(4.8164, device='cuda:0', grad_fn=<NllLossBackward0>) 149000 41700 tensor(5.1542, device='cuda:0', grad_fn=<NllLossBackward0>) 41800 tensor(5.0889, device='cuda:0', grad_fn=<NllLossBackward0>) 41900 tensor(5.0844, device='cuda:0', grad_fn=<NllLossBackward0>) 150000 42000 tensor(5.2015, device='cuda:0', grad_fn=<NllLossBackward0>) 42100 tensor(4.6232, device='cuda:0', grad_fn=<NllLossBackward0>) 151000 42200 tensor(4.9591, device='cuda:0', grad_fn=<NllLossBackward0>) 42300 tensor(5.0888, device='cuda:0', grad_fn=<NllLossBackward0>) 42400 tensor(4.9506, device='cuda:0', grad_fn=<NllLossBackward0>) 152000 42500 tensor(5.0405, device='cuda:0', grad_fn=<NllLossBackward0>) 42600 tensor(4.5948, device='cuda:0', grad_fn=<NllLossBackward0>) 42700 tensor(5.2494, device='cuda:0', grad_fn=<NllLossBackward0>) 153000 42800 tensor(5.1611, device='cuda:0', grad_fn=<NllLossBackward0>) 42900 tensor(4.9517, device='cuda:0', grad_fn=<NllLossBackward0>) 43000 tensor(5.1292, device='cuda:0', grad_fn=<NllLossBackward0>) 154000 43100 tensor(4.8461, device='cuda:0', grad_fn=<NllLossBackward0>) 43200 tensor(5.0209, device='cuda:0', grad_fn=<NllLossBackward0>) 43300 tensor(5.1240, device='cuda:0', grad_fn=<NllLossBackward0>) 155000 43400 tensor(5.0400, device='cuda:0', grad_fn=<NllLossBackward0>) 43500 tensor(5.3999, device='cuda:0', grad_fn=<NllLossBackward0>) 156000 43600 tensor(5.3041, device='cuda:0', grad_fn=<NllLossBackward0>) 43700 tensor(4.9734, device='cuda:0', grad_fn=<NllLossBackward0>) 43800 tensor(5.1028, device='cuda:0', grad_fn=<NllLossBackward0>) 157000 43900 tensor(4.9373, device='cuda:0', grad_fn=<NllLossBackward0>) 44000 tensor(4.7666, device='cuda:0', grad_fn=<NllLossBackward0>) 44100 tensor(5.2840, device='cuda:0', grad_fn=<NllLossBackward0>) 158000 44200 tensor(4.8623, device='cuda:0', grad_fn=<NllLossBackward0>) 44300 tensor(5.3904, device='cuda:0', grad_fn=<NllLossBackward0>) 44400 tensor(4.9841, device='cuda:0', grad_fn=<NllLossBackward0>) 159000 44500 tensor(4.8345, device='cuda:0', grad_fn=<NllLossBackward0>) 44600 tensor(4.9179, device='cuda:0', grad_fn=<NllLossBackward0>) 160000 44700 tensor(4.8921, device='cuda:0', grad_fn=<NllLossBackward0>) 44800 tensor(5.1909, device='cuda:0', grad_fn=<NllLossBackward0>) 44900 tensor(5.0083, device='cuda:0', grad_fn=<NllLossBackward0>) 161000 45000 tensor(4.8715, device='cuda:0', grad_fn=<NllLossBackward0>) 45100 tensor(5.1750, device='cuda:0', grad_fn=<NllLossBackward0>) 45200 tensor(4.9847, device='cuda:0', grad_fn=<NllLossBackward0>) 162000 45300 tensor(5.1416, device='cuda:0', grad_fn=<NllLossBackward0>) 45400 tensor(4.5706, device='cuda:0', grad_fn=<NllLossBackward0>) 45500 tensor(4.8476, device='cuda:0', grad_fn=<NllLossBackward0>) 163000 45600 tensor(5.1753, device='cuda:0', grad_fn=<NllLossBackward0>) 45700 tensor(5.1638, device='cuda:0', grad_fn=<NllLossBackward0>) 45800 tensor(5.0620, device='cuda:0', grad_fn=<NllLossBackward0>) 164000 45900 tensor(4.8623, device='cuda:0', grad_fn=<NllLossBackward0>) 46000 tensor(5.0958, device='cuda:0', grad_fn=<NllLossBackward0>) 165000 46100 tensor(4.9271, device='cuda:0', grad_fn=<NllLossBackward0>) 46200 tensor(4.8955, device='cuda:0', grad_fn=<NllLossBackward0>) 46300 tensor(5.2268, device='cuda:0', grad_fn=<NllLossBackward0>) 166000 46400 tensor(5.2585, device='cuda:0', grad_fn=<NllLossBackward0>) 46500 tensor(4.8509, device='cuda:0', grad_fn=<NllLossBackward0>) 46600 tensor(4.8656, device='cuda:0', grad_fn=<NllLossBackward0>) 167000 46700 tensor(5.1203, device='cuda:0', grad_fn=<NllLossBackward0>) 46800 tensor(4.7754, device='cuda:0', grad_fn=<NllLossBackward0>) 46900 tensor(4.8504, device='cuda:0', grad_fn=<NllLossBackward0>) 168000 47000 tensor(5.1042, device='cuda:0', grad_fn=<NllLossBackward0>) 47100 tensor(5.3206, device='cuda:0', grad_fn=<NllLossBackward0>) 47200 tensor(5.0201, device='cuda:0', grad_fn=<NllLossBackward0>) 169000 47300 tensor(5.1387, device='cuda:0', grad_fn=<NllLossBackward0>) 47400 tensor(4.9378, device='cuda:0', grad_fn=<NllLossBackward0>) 170000 47500 tensor(5.1120, device='cuda:0', grad_fn=<NllLossBackward0>) 47600 tensor(5.0451, device='cuda:0', grad_fn=<NllLossBackward0>) 47700 tensor(5.0578, device='cuda:0', grad_fn=<NllLossBackward0>) 171000 47800 tensor(5.1534, device='cuda:0', grad_fn=<NllLossBackward0>) 47900 tensor(5.1646, device='cuda:0', grad_fn=<NllLossBackward0>) 48000 tensor(4.8708, device='cuda:0', grad_fn=<NllLossBackward0>) 172000 48100 tensor(5.2243, device='cuda:0', grad_fn=<NllLossBackward0>) 48200 tensor(5.0376, device='cuda:0', grad_fn=<NllLossBackward0>) 48300 tensor(5.1148, device='cuda:0', grad_fn=<NllLossBackward0>) 173000 48400 tensor(4.7304, device='cuda:0', grad_fn=<NllLossBackward0>) 48500 tensor(4.9864, device='cuda:0', grad_fn=<NllLossBackward0>) 48600 tensor(5.1053, device='cuda:0', grad_fn=<NllLossBackward0>) 174000 48700 tensor(5.0854, device='cuda:0', grad_fn=<NllLossBackward0>) 48800 tensor(4.8705, device='cuda:0', grad_fn=<NllLossBackward0>) 175000 48900 tensor(5.1415, device='cuda:0', grad_fn=<NllLossBackward0>) 49000 tensor(4.9185, device='cuda:0', grad_fn=<NllLossBackward0>) 49100 tensor(5.1249, device='cuda:0', grad_fn=<NllLossBackward0>) 176000 49200 tensor(5.1852, device='cuda:0', grad_fn=<NllLossBackward0>) 49300 tensor(4.9728, device='cuda:0', grad_fn=<NllLossBackward0>) 49400 tensor(5.1541, device='cuda:0', grad_fn=<NllLossBackward0>) 177000 49500 tensor(5.2388, device='cuda:0', grad_fn=<NllLossBackward0>) 49600 tensor(4.8876, device='cuda:0', grad_fn=<NllLossBackward0>) 49700 tensor(5.0066, device='cuda:0', grad_fn=<NllLossBackward0>) 178000 49800 tensor(4.9596, device='cuda:0', grad_fn=<NllLossBackward0>) 49900 tensor(5.1618, device='cuda:0', grad_fn=<NllLossBackward0>) 50000 tensor(5.0436, device='cuda:0', grad_fn=<NllLossBackward0>) 179000 50100 tensor(5.1911, device='cuda:0', grad_fn=<NllLossBackward0>) 50200 tensor(4.9044, device='cuda:0', grad_fn=<NllLossBackward0>) 180000 50300 tensor(5.1698, device='cuda:0', grad_fn=<NllLossBackward0>) 50400 tensor(4.8785, device='cuda:0', grad_fn=<NllLossBackward0>) 50500 tensor(5.1889, device='cuda:0', grad_fn=<NllLossBackward0>) 181000 50600 tensor(4.9799, device='cuda:0', grad_fn=<NllLossBackward0>) 50700 tensor(4.8507, device='cuda:0', grad_fn=<NllLossBackward0>) 50800 tensor(4.7525, device='cuda:0', grad_fn=<NllLossBackward0>) 182000 50900 tensor(5.0997, device='cuda:0', grad_fn=<NllLossBackward0>) 51000 tensor(5.1681, device='cuda:0', grad_fn=<NllLossBackward0>) 51100 tensor(5.0502, device='cuda:0', grad_fn=<NllLossBackward0>) 183000 51200 tensor(5.2121, device='cuda:0', grad_fn=<NllLossBackward0>) 51300 tensor(4.7797, device='cuda:0', grad_fn=<NllLossBackward0>) 51400 tensor(5.1379, device='cuda:0', grad_fn=<NllLossBackward0>) 184000 51500 tensor(5.0826, device='cuda:0', grad_fn=<NllLossBackward0>) 51600 tensor(5.1736, device='cuda:0', grad_fn=<NllLossBackward0>) 185000 51700 tensor(5.0336, device='cuda:0', grad_fn=<NllLossBackward0>) 51800 tensor(5.1728, device='cuda:0', grad_fn=<NllLossBackward0>) 51900 tensor(4.9617, device='cuda:0', grad_fn=<NllLossBackward0>) 186000 52000 tensor(5.1423, device='cuda:0', grad_fn=<NllLossBackward0>) 52100 tensor(5.2641, device='cuda:0', grad_fn=<NllLossBackward0>) 52200 tensor(5.0587, device='cuda:0', grad_fn=<NllLossBackward0>) 187000 52300 tensor(4.9935, device='cuda:0', grad_fn=<NllLossBackward0>) 52400 tensor(5.0743, device='cuda:0', grad_fn=<NllLossBackward0>) 52500 tensor(5.0771, device='cuda:0', grad_fn=<NllLossBackward0>) 188000 52600 tensor(5.0109, device='cuda:0', grad_fn=<NllLossBackward0>) 52700 tensor(5.1416, device='cuda:0', grad_fn=<NllLossBackward0>) 189000 52800 tensor(5.1943, device='cuda:0', grad_fn=<NllLossBackward0>) 52900 tensor(5.0125, device='cuda:0', grad_fn=<NllLossBackward0>) 53000 tensor(4.9937, device='cuda:0', grad_fn=<NllLossBackward0>) 190000 53100 tensor(4.9477, device='cuda:0', grad_fn=<NllLossBackward0>) 53200 tensor(4.7327, device='cuda:0', grad_fn=<NllLossBackward0>) 53300 tensor(5.2251, device='cuda:0', grad_fn=<NllLossBackward0>) 191000 53400 tensor(5.1211, device='cuda:0', grad_fn=<NllLossBackward0>) 53500 tensor(5.0256, device='cuda:0', grad_fn=<NllLossBackward0>) 53600 tensor(5.1351, device='cuda:0', grad_fn=<NllLossBackward0>) 192000 53700 tensor(5.1121, device='cuda:0', grad_fn=<NllLossBackward0>) 53800 tensor(4.7721, device='cuda:0', grad_fn=<NllLossBackward0>) 53900 tensor(4.7813, device='cuda:0', grad_fn=<NllLossBackward0>) 193000 54000 tensor(5.2923, device='cuda:0', grad_fn=<NllLossBackward0>) 54100 tensor(4.4067, device='cuda:0', grad_fn=<NllLossBackward0>) 194000 54200 tensor(5.1834, device='cuda:0', grad_fn=<NllLossBackward0>) 54300 tensor(4.9698, device='cuda:0', grad_fn=<NllLossBackward0>) 54400 tensor(4.6432, device='cuda:0', grad_fn=<NllLossBackward0>) 195000 54500 tensor(5.2431, device='cuda:0', grad_fn=<NllLossBackward0>) 54600 tensor(5.2523, device='cuda:0', grad_fn=<NllLossBackward0>) 54700 tensor(4.9880, device='cuda:0', grad_fn=<NllLossBackward0>) 196000 54800 tensor(5.1705, device='cuda:0', grad_fn=<NllLossBackward0>) 54900 tensor(4.8544, device='cuda:0', grad_fn=<NllLossBackward0>) 55000 tensor(5.0054, device='cuda:0', grad_fn=<NllLossBackward0>) 197000 55100 tensor(5.0684, device='cuda:0', grad_fn=<NllLossBackward0>) 55200 tensor(5.1112, device='cuda:0', grad_fn=<NllLossBackward0>) 55300 tensor(5.0787, device='cuda:0', grad_fn=<NllLossBackward0>) 198000 55400 tensor(5.2173, device='cuda:0', grad_fn=<NllLossBackward0>) 55500 tensor(5.2430, device='cuda:0', grad_fn=<NllLossBackward0>) 199000 55600 tensor(4.8266, device='cuda:0', grad_fn=<NllLossBackward0>) 55700 tensor(5.0219, device='cuda:0', grad_fn=<NllLossBackward0>) 55800 tensor(4.5834, device='cuda:0', grad_fn=<NllLossBackward0>) 200000 55900 tensor(5.0455, device='cuda:0', grad_fn=<NllLossBackward0>) 56000 tensor(4.9394, device='cuda:0', grad_fn=<NllLossBackward0>) 56100 tensor(5.2400, device='cuda:0', grad_fn=<NllLossBackward0>) 201000 56200 tensor(5.2727, device='cuda:0', grad_fn=<NllLossBackward0>) 56300 tensor(5.0803, device='cuda:0', grad_fn=<NllLossBackward0>) 56400 tensor(5.1270, device='cuda:0', grad_fn=<NllLossBackward0>) 202000 56500 tensor(5.1758, device='cuda:0', grad_fn=<NllLossBackward0>) 56600 tensor(4.8928, device='cuda:0', grad_fn=<NllLossBackward0>) 56700 tensor(5.0119, device='cuda:0', grad_fn=<NllLossBackward0>) 203000 56800 tensor(5.5221, device='cuda:0', grad_fn=<NllLossBackward0>) 56900 tensor(4.8963, device='cuda:0', grad_fn=<NllLossBackward0>) 204000 57000 tensor(5.2899, device='cuda:0', grad_fn=<NllLossBackward0>) 57100 tensor(5.0742, device='cuda:0', grad_fn=<NllLossBackward0>) 57200 tensor(4.8940, device='cuda:0', grad_fn=<NllLossBackward0>) 205000 57300 tensor(5.0177, device='cuda:0', grad_fn=<NllLossBackward0>) 57400 tensor(5.2162, device='cuda:0', grad_fn=<NllLossBackward0>) 57500 tensor(4.8549, device='cuda:0', grad_fn=<NllLossBackward0>) 206000 57600 tensor(4.7256, device='cuda:0', grad_fn=<NllLossBackward0>) 57700 tensor(4.9188, device='cuda:0', grad_fn=<NllLossBackward0>) 57800 tensor(5.2799, device='cuda:0', grad_fn=<NllLossBackward0>) 207000 57900 tensor(5.2005, device='cuda:0', grad_fn=<NllLossBackward0>) 58000 tensor(4.9628, device='cuda:0', grad_fn=<NllLossBackward0>) 208000 58100 tensor(5.3107, device='cuda:0', grad_fn=<NllLossBackward0>) 58200 tensor(5.3371, device='cuda:0', grad_fn=<NllLossBackward0>) 58300 tensor(4.9034, device='cuda:0', grad_fn=<NllLossBackward0>) 209000 58400 tensor(4.8145, device='cuda:0', grad_fn=<NllLossBackward0>) 58500 tensor(4.9395, device='cuda:0', grad_fn=<NllLossBackward0>) 58600 tensor(5.3965, device='cuda:0', grad_fn=<NllLossBackward0>) 210000 58700 tensor(4.9055, device='cuda:0', grad_fn=<NllLossBackward0>) 58800 tensor(5.0192, device='cuda:0', grad_fn=<NllLossBackward0>) 58900 tensor(5.2319, device='cuda:0', grad_fn=<NllLossBackward0>) 211000 59000 tensor(5.1120, device='cuda:0', grad_fn=<NllLossBackward0>) 59100 tensor(4.9841, device='cuda:0', grad_fn=<NllLossBackward0>) 59200 tensor(5.0341, device='cuda:0', grad_fn=<NllLossBackward0>) 212000 59300 tensor(4.8836, device='cuda:0', grad_fn=<NllLossBackward0>) 59400 tensor(5.0700, device='cuda:0', grad_fn=<NllLossBackward0>) 213000 59500 tensor(5.0094, device='cuda:0', grad_fn=<NllLossBackward0>) 59600 tensor(4.9719, device='cuda:0', grad_fn=<NllLossBackward0>) 59700 tensor(4.8263, device='cuda:0', grad_fn=<NllLossBackward0>) 214000 59800 tensor(5.0422, device='cuda:0', grad_fn=<NllLossBackward0>) 59900 tensor(5.0391, device='cuda:0', grad_fn=<NllLossBackward0>) 60000 tensor(5.0464, device='cuda:0', grad_fn=<NllLossBackward0>) 215000 60100 tensor(5.1302, device='cuda:0', grad_fn=<NllLossBackward0>) 60200 tensor(5.1667, device='cuda:0', grad_fn=<NllLossBackward0>) 60300 tensor(4.9755, device='cuda:0', grad_fn=<NllLossBackward0>) 216000 60400 tensor(5.0662, device='cuda:0', grad_fn=<NllLossBackward0>) 60500 tensor(5.1330, device='cuda:0', grad_fn=<NllLossBackward0>) 60600 tensor(5.0362, device='cuda:0', grad_fn=<NllLossBackward0>) 217000 60700 tensor(4.9462, device='cuda:0', grad_fn=<NllLossBackward0>) 60800 tensor(5.1028, device='cuda:0', grad_fn=<NllLossBackward0>) 218000 60900 tensor(5.1816, device='cuda:0', grad_fn=<NllLossBackward0>) 61000 tensor(5.1451, device='cuda:0', grad_fn=<NllLossBackward0>) 61100 tensor(5.0389, device='cuda:0', grad_fn=<NllLossBackward0>) 219000 61200 tensor(4.7783, device='cuda:0', grad_fn=<NllLossBackward0>) 61300 tensor(5.0208, device='cuda:0', grad_fn=<NllLossBackward0>) 61400 tensor(5.2014, device='cuda:0', grad_fn=<NllLossBackward0>) 220000 61500 tensor(5.0953, device='cuda:0', grad_fn=<NllLossBackward0>) 61600 tensor(4.9255, device='cuda:0', grad_fn=<NllLossBackward0>) 61700 tensor(4.9145, device='cuda:0', grad_fn=<NllLossBackward0>) 221000 61800 tensor(4.9751, device='cuda:0', grad_fn=<NllLossBackward0>) 61900 tensor(5.0497, device='cuda:0', grad_fn=<NllLossBackward0>) 62000 tensor(5.2065, device='cuda:0', grad_fn=<NllLossBackward0>) 222000 62100 tensor(4.8016, device='cuda:0', grad_fn=<NllLossBackward0>) 62200 tensor(4.9131, device='cuda:0', grad_fn=<NllLossBackward0>) 223000 62300 tensor(4.9816, device='cuda:0', grad_fn=<NllLossBackward0>) 62400 tensor(4.9049, device='cuda:0', grad_fn=<NllLossBackward0>) 62500 tensor(5.0251, device='cuda:0', grad_fn=<NllLossBackward0>) 224000 62600 tensor(4.9373, device='cuda:0', grad_fn=<NllLossBackward0>) 62700 tensor(5.0984, device='cuda:0', grad_fn=<NllLossBackward0>) 62800 tensor(4.8395, device='cuda:0', grad_fn=<NllLossBackward0>) 225000 62900 tensor(5.3108, device='cuda:0', grad_fn=<NllLossBackward0>) 63000 tensor(5.1682, device='cuda:0', grad_fn=<NllLossBackward0>) 63100 tensor(5.2454, device='cuda:0', grad_fn=<NllLossBackward0>) 226000 63200 tensor(5.1091, device='cuda:0', grad_fn=<NllLossBackward0>) 63300 tensor(4.8577, device='cuda:0', grad_fn=<NllLossBackward0>) 63400 tensor(5.0426, device='cuda:0', grad_fn=<NllLossBackward0>) 227000 63500 tensor(4.9117, device='cuda:0', grad_fn=<NllLossBackward0>) 63600 tensor(4.9555, device='cuda:0', grad_fn=<NllLossBackward0>) 228000 63700 tensor(4.9914, device='cuda:0', grad_fn=<NllLossBackward0>) 63800 tensor(5.2633, device='cuda:0', grad_fn=<NllLossBackward0>) 63900 tensor(5.3451, device='cuda:0', grad_fn=<NllLossBackward0>) 229000 64000 tensor(4.9019, device='cuda:0', grad_fn=<NllLossBackward0>) 64100 tensor(5.1581, device='cuda:0', grad_fn=<NllLossBackward0>) 64200 tensor(5.0796, device='cuda:0', grad_fn=<NllLossBackward0>) 230000 64300 tensor(5.0816, device='cuda:0', grad_fn=<NllLossBackward0>) 64400 tensor(4.9050, device='cuda:0', grad_fn=<NllLossBackward0>) 64500 tensor(4.9127, device='cuda:0', grad_fn=<NllLossBackward0>) 231000 64600 tensor(4.9467, device='cuda:0', grad_fn=<NllLossBackward0>) 64700 tensor(5.3998, device='cuda:0', grad_fn=<NllLossBackward0>) 64800 tensor(5.0784, device='cuda:0', grad_fn=<NllLossBackward0>) 232000 64900 tensor(5.2337, device='cuda:0', grad_fn=<NllLossBackward0>) 65000 tensor(5.1862, device='cuda:0', grad_fn=<NllLossBackward0>) 233000 65100 tensor(4.7268, device='cuda:0', grad_fn=<NllLossBackward0>) 65200 tensor(5.2401, device='cuda:0', grad_fn=<NllLossBackward0>) 65300 tensor(5.0779, device='cuda:0', grad_fn=<NllLossBackward0>) 234000 65400 tensor(5.0225, device='cuda:0', grad_fn=<NllLossBackward0>) 65500 tensor(5.0218, device='cuda:0', grad_fn=<NllLossBackward0>) 65600 tensor(5.1051, device='cuda:0', grad_fn=<NllLossBackward0>) 235000 65700 tensor(4.8759, device='cuda:0', grad_fn=<NllLossBackward0>) 65800 tensor(4.9007, device='cuda:0', grad_fn=<NllLossBackward0>) 65900 tensor(5.1119, device='cuda:0', grad_fn=<NllLossBackward0>) 236000 66000 tensor(5.1512, device='cuda:0', grad_fn=<NllLossBackward0>) 66100 tensor(4.8563, device='cuda:0', grad_fn=<NllLossBackward0>) 237000 66200 tensor(4.8676, device='cuda:0', grad_fn=<NllLossBackward0>) 66300 tensor(4.8076, device='cuda:0', grad_fn=<NllLossBackward0>) 66400 tensor(5.0870, device='cuda:0', grad_fn=<NllLossBackward0>) 238000 66500 tensor(4.8053, device='cuda:0', grad_fn=<NllLossBackward0>) 66600 tensor(4.9573, device='cuda:0', grad_fn=<NllLossBackward0>) 66700 tensor(4.9532, device='cuda:0', grad_fn=<NllLossBackward0>) 239000 66800 tensor(5.0142, device='cuda:0', grad_fn=<NllLossBackward0>) 66900 tensor(5.1463, device='cuda:0', grad_fn=<NllLossBackward0>) 67000 tensor(5.1613, device='cuda:0', grad_fn=<NllLossBackward0>) 240000 67100 tensor(5.0125, device='cuda:0', grad_fn=<NllLossBackward0>) 67200 tensor(5.3587, device='cuda:0', grad_fn=<NllLossBackward0>) 67300 tensor(4.9968, device='cuda:0', grad_fn=<NllLossBackward0>) 241000 67400 tensor(5.1922, device='cuda:0', grad_fn=<NllLossBackward0>) 67500 tensor(5.0287, device='cuda:0', grad_fn=<NllLossBackward0>) 242000 67600 tensor(5.0606, device='cuda:0', grad_fn=<NllLossBackward0>) 67700 tensor(5.0029, device='cuda:0', grad_fn=<NllLossBackward0>) 67800 tensor(5.1391, device='cuda:0', grad_fn=<NllLossBackward0>) 243000 67900 tensor(4.8602, device='cuda:0', grad_fn=<NllLossBackward0>) 68000 tensor(4.9282, device='cuda:0', grad_fn=<NllLossBackward0>) 68100 tensor(4.7940, device='cuda:0', grad_fn=<NllLossBackward0>) 244000 68200 tensor(5.0718, device='cuda:0', grad_fn=<NllLossBackward0>) 68300 tensor(4.9727, device='cuda:0', grad_fn=<NllLossBackward0>) 68400 tensor(5.0790, device='cuda:0', grad_fn=<NllLossBackward0>) 245000 68500 tensor(5.0921, device='cuda:0', grad_fn=<NllLossBackward0>) 68600 tensor(5.0522, device='cuda:0', grad_fn=<NllLossBackward0>) 68700 tensor(4.7111, device='cuda:0', grad_fn=<NllLossBackward0>) 246000 68800 tensor(5.1136, device='cuda:0', grad_fn=<NllLossBackward0>) 68900 tensor(4.7838, device='cuda:0', grad_fn=<NllLossBackward0>) 247000 69000 tensor(5.0396, device='cuda:0', grad_fn=<NllLossBackward0>) 69100 tensor(4.8118, device='cuda:0', grad_fn=<NllLossBackward0>) 69200 tensor(4.7698, device='cuda:0', grad_fn=<NllLossBackward0>) 248000 69300 tensor(5.2231, device='cuda:0', grad_fn=<NllLossBackward0>) 69400 tensor(5.1113, device='cuda:0', grad_fn=<NllLossBackward0>) 69500 tensor(5.0799, device='cuda:0', grad_fn=<NllLossBackward0>) 249000 69600 tensor(5.0015, device='cuda:0', grad_fn=<NllLossBackward0>) 69700 tensor(5.3189, device='cuda:0', grad_fn=<NllLossBackward0>) 69800 tensor(4.8597, device='cuda:0', grad_fn=<NllLossBackward0>) 250000 69900 tensor(5.4064, device='cuda:0', grad_fn=<NllLossBackward0>) 70000 tensor(5.0562, device='cuda:0', grad_fn=<NllLossBackward0>) 70100 tensor(4.9788, device='cuda:0', grad_fn=<NllLossBackward0>) 251000 70200 tensor(5.1238, device='cuda:0', grad_fn=<NllLossBackward0>) 70300 tensor(4.7954, device='cuda:0', grad_fn=<NllLossBackward0>) 252000 70400 tensor(5.1417, device='cuda:0', grad_fn=<NllLossBackward0>) 70500 tensor(5.0851, device='cuda:0', grad_fn=<NllLossBackward0>) 70600 tensor(5.3102, device='cuda:0', grad_fn=<NllLossBackward0>) 253000 70700 tensor(5.2187, device='cuda:0', grad_fn=<NllLossBackward0>) 70800 tensor(5.4180, device='cuda:0', grad_fn=<NllLossBackward0>) 70900 tensor(5.2464, device='cuda:0', grad_fn=<NllLossBackward0>) 254000 71000 tensor(5.0232, device='cuda:0', grad_fn=<NllLossBackward0>) 71100 tensor(5.0971, device='cuda:0', grad_fn=<NllLossBackward0>) 71200 tensor(5.2289, device='cuda:0', grad_fn=<NllLossBackward0>) 255000 71300 tensor(4.8301, device='cuda:0', grad_fn=<NllLossBackward0>) 71400 tensor(4.9594, device='cuda:0', grad_fn=<NllLossBackward0>) 71500 tensor(4.9527, device='cuda:0', grad_fn=<NllLossBackward0>) 256000 71600 tensor(5.1395, device='cuda:0', grad_fn=<NllLossBackward0>) 71700 tensor(5.2029, device='cuda:0', grad_fn=<NllLossBackward0>) 257000 71800 tensor(5.2958, device='cuda:0', grad_fn=<NllLossBackward0>) 71900 tensor(4.9839, device='cuda:0', grad_fn=<NllLossBackward0>) 72000 tensor(5.4251, device='cuda:0', grad_fn=<NllLossBackward0>) 258000 72100 tensor(5.1480, device='cuda:0', grad_fn=<NllLossBackward0>) 72200 tensor(4.9666, device='cuda:0', grad_fn=<NllLossBackward0>) 72300 tensor(5.0831, device='cuda:0', grad_fn=<NllLossBackward0>) 259000 72400 tensor(4.9905, device='cuda:0', grad_fn=<NllLossBackward0>) 72500 tensor(5.0817, device='cuda:0', grad_fn=<NllLossBackward0>) 72600 tensor(5.1276, device='cuda:0', grad_fn=<NllLossBackward0>) 260000 72700 tensor(4.8184, device='cuda:0', grad_fn=<NllLossBackward0>) 72800 tensor(5.2583, device='cuda:0', grad_fn=<NllLossBackward0>) 72900 tensor(4.9964, device='cuda:0', grad_fn=<NllLossBackward0>) 261000 73000 tensor(5.1877, device='cuda:0', grad_fn=<NllLossBackward0>) 73100 tensor(4.8954, device='cuda:0', grad_fn=<NllLossBackward0>) 262000 73200 tensor(4.9722, device='cuda:0', grad_fn=<NllLossBackward0>) 73300 tensor(4.9351, device='cuda:0', grad_fn=<NllLossBackward0>) 73400 tensor(5.2009, device='cuda:0', grad_fn=<NllLossBackward0>) 263000 73500 tensor(5.0204, device='cuda:0', grad_fn=<NllLossBackward0>) 73600 tensor(4.9348, device='cuda:0', grad_fn=<NllLossBackward0>) 73700 tensor(5.0774, device='cuda:0', grad_fn=<NllLossBackward0>) 264000 73800 tensor(5.1100, device='cuda:0', grad_fn=<NllLossBackward0>) 73900 tensor(5.1179, device='cuda:0', grad_fn=<NllLossBackward0>) 74000 tensor(5.2795, device='cuda:0', grad_fn=<NllLossBackward0>) 265000 74100 tensor(4.9642, device='cuda:0', grad_fn=<NllLossBackward0>) 74200 tensor(4.9902, device='cuda:0', grad_fn=<NllLossBackward0>) 74300 tensor(4.9494, device='cuda:0', grad_fn=<NllLossBackward0>) 266000 74400 tensor(5.0772, device='cuda:0', grad_fn=<NllLossBackward0>) 74500 tensor(4.6984, device='cuda:0', grad_fn=<NllLossBackward0>) 267000 74600 tensor(5.2123, device='cuda:0', grad_fn=<NllLossBackward0>) 74700 tensor(5.2695, device='cuda:0', grad_fn=<NllLossBackward0>) 74800 tensor(4.9678, device='cuda:0', grad_fn=<NllLossBackward0>) 268000 74900 tensor(4.7941, device='cuda:0', grad_fn=<NllLossBackward0>) 75000 tensor(5.0980, device='cuda:0', grad_fn=<NllLossBackward0>) 75100 tensor(5.2916, device='cuda:0', grad_fn=<NllLossBackward0>) 269000 75200 tensor(5.0355, device='cuda:0', grad_fn=<NllLossBackward0>) 75300 tensor(5.2583, device='cuda:0', grad_fn=<NllLossBackward0>) 75400 tensor(5.0347, device='cuda:0', grad_fn=<NllLossBackward0>) 270000 75500 tensor(5.3012, device='cuda:0', grad_fn=<NllLossBackward0>) 75600 tensor(4.9714, device='cuda:0', grad_fn=<NllLossBackward0>) 271000 75700 tensor(5.0178, device='cuda:0', grad_fn=<NllLossBackward0>) 75800 tensor(4.9899, device='cuda:0', grad_fn=<NllLossBackward0>) 75900 tensor(5.0695, device='cuda:0', grad_fn=<NllLossBackward0>) 272000 76000 tensor(4.8497, device='cuda:0', grad_fn=<NllLossBackward0>) 76100 tensor(4.8143, device='cuda:0', grad_fn=<NllLossBackward0>) 76200 tensor(4.8724, device='cuda:0', grad_fn=<NllLossBackward0>) 273000 76300 tensor(5.0268, device='cuda:0', grad_fn=<NllLossBackward0>) 76400 tensor(4.8836, device='cuda:0', grad_fn=<NllLossBackward0>) 76500 tensor(5.0496, device='cuda:0', grad_fn=<NllLossBackward0>) 274000 76600 tensor(5.1337, device='cuda:0', grad_fn=<NllLossBackward0>) 76700 tensor(4.8128, device='cuda:0', grad_fn=<NllLossBackward0>) 76800 tensor(4.7389, device='cuda:0', grad_fn=<NllLossBackward0>) 275000 76900 tensor(5.0926, device='cuda:0', grad_fn=<NllLossBackward0>) 77000 tensor(4.9780, device='cuda:0', grad_fn=<NllLossBackward0>) 276000 77100 tensor(5.3241, device='cuda:0', grad_fn=<NllLossBackward0>) 77200 tensor(4.9880, device='cuda:0', grad_fn=<NllLossBackward0>) 77300 tensor(4.9250, device='cuda:0', grad_fn=<NllLossBackward0>) 277000 77400 tensor(5.0480, device='cuda:0', grad_fn=<NllLossBackward0>) 77500 tensor(4.7045, device='cuda:0', grad_fn=<NllLossBackward0>) 77600 tensor(5.0635, device='cuda:0', grad_fn=<NllLossBackward0>) 278000 77700 tensor(5.2122, device='cuda:0', grad_fn=<NllLossBackward0>) 77800 tensor(4.8158, device='cuda:0', grad_fn=<NllLossBackward0>) 77900 tensor(4.7201, device='cuda:0', grad_fn=<NllLossBackward0>) 279000 78000 tensor(5.1809, device='cuda:0', grad_fn=<NllLossBackward0>) 78100 tensor(4.8416, device='cuda:0', grad_fn=<NllLossBackward0>) 78200 tensor(5.1227, device='cuda:0', grad_fn=<NllLossBackward0>) 280000 78300 tensor(5.0000, device='cuda:0', grad_fn=<NllLossBackward0>) 78400 tensor(5.0232, device='cuda:0', grad_fn=<NllLossBackward0>) 281000 78500 tensor(5.3672, device='cuda:0', grad_fn=<NllLossBackward0>) 78600 tensor(4.8266, device='cuda:0', grad_fn=<NllLossBackward0>) 78700 tensor(5.2883, device='cuda:0', grad_fn=<NllLossBackward0>) 282000 78800 tensor(5.1556, device='cuda:0', grad_fn=<NllLossBackward0>) 78900 tensor(5.3901, device='cuda:0', grad_fn=<NllLossBackward0>) 79000 tensor(4.9368, device='cuda:0', grad_fn=<NllLossBackward0>) 283000 79100 tensor(4.7262, device='cuda:0', grad_fn=<NllLossBackward0>) 79200 tensor(4.8004, device='cuda:0', grad_fn=<NllLossBackward0>) 79300 tensor(5.2785, device='cuda:0', grad_fn=<NllLossBackward0>) 284000 79400 tensor(5.0932, device='cuda:0', grad_fn=<NllLossBackward0>) 79500 tensor(4.9747, device='cuda:0', grad_fn=<NllLossBackward0>) 79600 tensor(5.5676, device='cuda:0', grad_fn=<NllLossBackward0>) 285000 79700 tensor(4.7581, device='cuda:0', grad_fn=<NllLossBackward0>) 79800 tensor(5.0427, device='cuda:0', grad_fn=<NllLossBackward0>) 286000 79900 tensor(5.0180, device='cuda:0', grad_fn=<NllLossBackward0>) 80000 tensor(5.1512, device='cuda:0', grad_fn=<NllLossBackward0>) 80100 tensor(5.2409, device='cuda:0', grad_fn=<NllLossBackward0>) 287000 80200 tensor(5.2435, device='cuda:0', grad_fn=<NllLossBackward0>) 80300 tensor(5.2283, device='cuda:0', grad_fn=<NllLossBackward0>) 80400 tensor(5.3199, device='cuda:0', grad_fn=<NllLossBackward0>) 288000 80500 tensor(5.0696, device='cuda:0', grad_fn=<NllLossBackward0>) 80600 tensor(5.2085, device='cuda:0', grad_fn=<NllLossBackward0>) 80700 tensor(5.2603, device='cuda:0', grad_fn=<NllLossBackward0>) 289000 80800 tensor(4.7770, device='cuda:0', grad_fn=<NllLossBackward0>) 80900 tensor(5.1645, device='cuda:0', grad_fn=<NllLossBackward0>) 290000 81000 tensor(5.1184, device='cuda:0', grad_fn=<NllLossBackward0>) 81100 tensor(5.1388, device='cuda:0', grad_fn=<NllLossBackward0>) 81200 tensor(4.8794, device='cuda:0', grad_fn=<NllLossBackward0>) 291000 81300 tensor(4.9514, device='cuda:0', grad_fn=<NllLossBackward0>) 81400 tensor(5.0975, device='cuda:0', grad_fn=<NllLossBackward0>) 81500 tensor(5.0673, device='cuda:0', grad_fn=<NllLossBackward0>) 292000 81600 tensor(5.0619, device='cuda:0', grad_fn=<NllLossBackward0>) 81700 tensor(5.1551, device='cuda:0', grad_fn=<NllLossBackward0>) 81800 tensor(5.0009, device='cuda:0', grad_fn=<NllLossBackward0>) 293000 81900 tensor(5.2308, device='cuda:0', grad_fn=<NllLossBackward0>) 82000 tensor(5.0541, device='cuda:0', grad_fn=<NllLossBackward0>) 82100 tensor(5.0817, device='cuda:0', grad_fn=<NllLossBackward0>) 294000 82200 tensor(5.0208, device='cuda:0', grad_fn=<NllLossBackward0>) 82300 tensor(4.8939, device='cuda:0', grad_fn=<NllLossBackward0>) 295000 82400 tensor(5.1285, device='cuda:0', grad_fn=<NllLossBackward0>) 82500 tensor(4.8489, device='cuda:0', grad_fn=<NllLossBackward0>) 82600 tensor(5.0365, device='cuda:0', grad_fn=<NllLossBackward0>) 296000 82700 tensor(5.0726, device='cuda:0', grad_fn=<NllLossBackward0>) 82800 tensor(5.0563, device='cuda:0', grad_fn=<NllLossBackward0>) 82900 tensor(4.7439, device='cuda:0', grad_fn=<NllLossBackward0>) 297000 83000 tensor(5.1423, device='cuda:0', grad_fn=<NllLossBackward0>) 83100 tensor(5.0052, device='cuda:0', grad_fn=<NllLossBackward0>) 83200 tensor(5.2631, device='cuda:0', grad_fn=<NllLossBackward0>) 298000 83300 tensor(4.9471, device='cuda:0', grad_fn=<NllLossBackward0>) 83400 tensor(5.0234, device='cuda:0', grad_fn=<NllLossBackward0>) 83500 tensor(4.9797, device='cuda:0', grad_fn=<NllLossBackward0>) 299000 83600 tensor(4.9047, device='cuda:0', grad_fn=<NllLossBackward0>) 83700 tensor(5.1383, device='cuda:0', grad_fn=<NllLossBackward0>) 300000 83800 tensor(4.8484, device='cuda:0', grad_fn=<NllLossBackward0>) 83900 tensor(4.9279, device='cuda:0', grad_fn=<NllLossBackward0>) 84000 tensor(5.0825, device='cuda:0', grad_fn=<NllLossBackward0>) 301000 84100 tensor(5.4376, device='cuda:0', grad_fn=<NllLossBackward0>) 84200 tensor(4.5984, device='cuda:0', grad_fn=<NllLossBackward0>) 84300 tensor(5.0169, device='cuda:0', grad_fn=<NllLossBackward0>) 302000 84400 tensor(4.8503, device='cuda:0', grad_fn=<NllLossBackward0>) 84500 tensor(5.0322, device='cuda:0', grad_fn=<NllLossBackward0>) 84600 tensor(5.2033, device='cuda:0', grad_fn=<NllLossBackward0>) 303000 84700 tensor(4.9052, device='cuda:0', grad_fn=<NllLossBackward0>) 84800 tensor(4.9265, device='cuda:0', grad_fn=<NllLossBackward0>) 84900 tensor(5.2419, device='cuda:0', grad_fn=<NllLossBackward0>) 304000 85000 tensor(5.1251, device='cuda:0', grad_fn=<NllLossBackward0>) 85100 tensor(5.0005, device='cuda:0', grad_fn=<NllLossBackward0>) 305000 85200 tensor(5.0029, device='cuda:0', grad_fn=<NllLossBackward0>) 85300 tensor(5.2717, device='cuda:0', grad_fn=<NllLossBackward0>) 85400 tensor(5.2511, device='cuda:0', grad_fn=<NllLossBackward0>) 306000 85500 tensor(4.9489, device='cuda:0', grad_fn=<NllLossBackward0>) 85600 tensor(5.0751, device='cuda:0', grad_fn=<NllLossBackward0>) 85700 tensor(5.0649, device='cuda:0', grad_fn=<NllLossBackward0>) 307000 85800 tensor(5.0660, device='cuda:0', grad_fn=<NllLossBackward0>) 85900 tensor(5.3243, device='cuda:0', grad_fn=<NllLossBackward0>) 86000 tensor(5.0756, device='cuda:0', grad_fn=<NllLossBackward0>) 308000 86100 tensor(4.8535, device='cuda:0', grad_fn=<NllLossBackward0>) 86200 tensor(4.9982, device='cuda:0', grad_fn=<NllLossBackward0>) 86300 tensor(4.9693, device='cuda:0', grad_fn=<NllLossBackward0>) 309000 86400 tensor(5.0700, device='cuda:0', grad_fn=<NllLossBackward0>) 86500 tensor(5.1470, device='cuda:0', grad_fn=<NllLossBackward0>) 310000 86600 tensor(4.9735, device='cuda:0', grad_fn=<NllLossBackward0>) 86700 tensor(4.8743, device='cuda:0', grad_fn=<NllLossBackward0>) 86800 tensor(4.8816, device='cuda:0', grad_fn=<NllLossBackward0>) 311000 86900 tensor(4.7666, device='cuda:0', grad_fn=<NllLossBackward0>) 87000 tensor(5.2600, device='cuda:0', grad_fn=<NllLossBackward0>) 87100 tensor(5.0591, device='cuda:0', grad_fn=<NllLossBackward0>) 312000 87200 tensor(5.1044, device='cuda:0', grad_fn=<NllLossBackward0>) 87300 tensor(4.9555, device='cuda:0', grad_fn=<NllLossBackward0>) 87400 tensor(5.0028, device='cuda:0', grad_fn=<NllLossBackward0>) 313000 87500 tensor(5.1317, device='cuda:0', grad_fn=<NllLossBackward0>) 87600 tensor(5.2621, device='cuda:0', grad_fn=<NllLossBackward0>) 87700 tensor(4.8937, device='cuda:0', grad_fn=<NllLossBackward0>) 314000 87800 tensor(4.8559, device='cuda:0', grad_fn=<NllLossBackward0>) 87900 tensor(5.1339, device='cuda:0', grad_fn=<NllLossBackward0>) 315000 88000 tensor(4.6132, device='cuda:0', grad_fn=<NllLossBackward0>) 88100 tensor(4.9682, device='cuda:0', grad_fn=<NllLossBackward0>) 88200 tensor(5.1715, device='cuda:0', grad_fn=<NllLossBackward0>) 316000 88300 tensor(5.1112, device='cuda:0', grad_fn=<NllLossBackward0>) 88400 tensor(5.1811, device='cuda:0', grad_fn=<NllLossBackward0>) 88500 tensor(4.8046, device='cuda:0', grad_fn=<NllLossBackward0>) 317000 88600 tensor(4.8210, device='cuda:0', grad_fn=<NllLossBackward0>) 88700 tensor(4.8294, device='cuda:0', grad_fn=<NllLossBackward0>) 88800 tensor(4.8198, device='cuda:0', grad_fn=<NllLossBackward0>) 318000 88900 tensor(4.7496, device='cuda:0', grad_fn=<NllLossBackward0>) 89000 tensor(4.5654, device='cuda:0', grad_fn=<NllLossBackward0>) 89100 tensor(5.1261, device='cuda:0', grad_fn=<NllLossBackward0>) 319000 89200 tensor(4.9374, device='cuda:0', grad_fn=<NllLossBackward0>) 89300 tensor(5.3199, device='cuda:0', grad_fn=<NllLossBackward0>) 320000 89400 tensor(4.6848, device='cuda:0', grad_fn=<NllLossBackward0>) 89500 tensor(5.0381, device='cuda:0', grad_fn=<NllLossBackward0>) 89600 tensor(5.2632, device='cuda:0', grad_fn=<NllLossBackward0>) 321000 89700 tensor(4.8221, device='cuda:0', grad_fn=<NllLossBackward0>) 89800 tensor(5.0413, device='cuda:0', grad_fn=<NllLossBackward0>) 89900 tensor(5.0402, device='cuda:0', grad_fn=<NllLossBackward0>) 322000 90000 tensor(5.0519, device='cuda:0', grad_fn=<NllLossBackward0>) 90100 tensor(4.9362, device='cuda:0', grad_fn=<NllLossBackward0>) 90200 tensor(5.0257, device='cuda:0', grad_fn=<NllLossBackward0>) 323000 90300 tensor(5.0438, device='cuda:0', grad_fn=<NllLossBackward0>) 90400 tensor(4.7173, device='cuda:0', grad_fn=<NllLossBackward0>) 90500 tensor(4.9121, device='cuda:0', grad_fn=<NllLossBackward0>) 324000 90600 tensor(5.1965, device='cuda:0', grad_fn=<NllLossBackward0>) 90700 tensor(4.8207, device='cuda:0', grad_fn=<NllLossBackward0>) 325000 90800 tensor(5.0536, device='cuda:0', grad_fn=<NllLossBackward0>) 90900 tensor(5.1080, device='cuda:0', grad_fn=<NllLossBackward0>) 91000 tensor(5.1328, device='cuda:0', grad_fn=<NllLossBackward0>) 326000 91100 tensor(4.7033, device='cuda:0', grad_fn=<NllLossBackward0>) 91200 tensor(5.1643, device='cuda:0', grad_fn=<NllLossBackward0>) 91300 tensor(5.3472, device='cuda:0', grad_fn=<NllLossBackward0>) 327000 91400 tensor(4.7267, device='cuda:0', grad_fn=<NllLossBackward0>) 91500 tensor(4.9552, device='cuda:0', grad_fn=<NllLossBackward0>) 91600 tensor(5.2054, device='cuda:0', grad_fn=<NllLossBackward0>) 328000 91700 tensor(5.0471, device='cuda:0', grad_fn=<NllLossBackward0>) 91800 tensor(5.2265, device='cuda:0', grad_fn=<NllLossBackward0>) 329000 91900 tensor(5.0523, device='cuda:0', grad_fn=<NllLossBackward0>) 92000 tensor(5.0726, device='cuda:0', grad_fn=<NllLossBackward0>) 92100 tensor(4.9322, device='cuda:0', grad_fn=<NllLossBackward0>) 330000 92200 tensor(4.5777, device='cuda:0', grad_fn=<NllLossBackward0>) 92300 tensor(5.0606, device='cuda:0', grad_fn=<NllLossBackward0>) 92400 tensor(5.1319, device='cuda:0', grad_fn=<NllLossBackward0>) 331000 92500 tensor(5.1785, device='cuda:0', grad_fn=<NllLossBackward0>) 92600 tensor(5.0610, device='cuda:0', grad_fn=<NllLossBackward0>) 92700 tensor(4.7718, device='cuda:0', grad_fn=<NllLossBackward0>) 332000 92800 tensor(4.8492, device='cuda:0', grad_fn=<NllLossBackward0>) 92900 tensor(5.3673, device='cuda:0', grad_fn=<NllLossBackward0>) 93000 tensor(5.0656, device='cuda:0', grad_fn=<NllLossBackward0>) 333000 93100 tensor(4.9975, device='cuda:0', grad_fn=<NllLossBackward0>) 93200 tensor(5.1829, device='cuda:0', grad_fn=<NllLossBackward0>) 334000 93300 tensor(5.1131, device='cuda:0', grad_fn=<NllLossBackward0>) 93400 tensor(5.1585, device='cuda:0', grad_fn=<NllLossBackward0>) 93500 tensor(4.9141, device='cuda:0', grad_fn=<NllLossBackward0>) 335000 93600 tensor(5.0526, device='cuda:0', grad_fn=<NllLossBackward0>) 93700 tensor(5.1195, device='cuda:0', grad_fn=<NllLossBackward0>) 93800 tensor(4.8696, device='cuda:0', grad_fn=<NllLossBackward0>) 336000 93900 tensor(5.0752, device='cuda:0', grad_fn=<NllLossBackward0>) 94000 tensor(4.9943, device='cuda:0', grad_fn=<NllLossBackward0>) 94100 tensor(4.8657, device='cuda:0', grad_fn=<NllLossBackward0>) 337000 94200 tensor(5.1230, device='cuda:0', grad_fn=<NllLossBackward0>) 94300 tensor(5.0292, device='cuda:0', grad_fn=<NllLossBackward0>) 94400 tensor(5.0633, device='cuda:0', grad_fn=<NllLossBackward0>) 338000 94500 tensor(4.9025, device='cuda:0', grad_fn=<NllLossBackward0>) 94600 tensor(5.1955, device='cuda:0', grad_fn=<NllLossBackward0>) 339000 94700 tensor(4.7546, device='cuda:0', grad_fn=<NllLossBackward0>) 94800 tensor(5.2260, device='cuda:0', grad_fn=<NllLossBackward0>) 94900 tensor(5.2268, device='cuda:0', grad_fn=<NllLossBackward0>) 340000 95000 tensor(5.3426, device='cuda:0', grad_fn=<NllLossBackward0>) 95100 tensor(5.0509, device='cuda:0', grad_fn=<NllLossBackward0>) 95200 tensor(4.8831, device='cuda:0', grad_fn=<NllLossBackward0>) 341000 95300 tensor(5.2464, device='cuda:0', grad_fn=<NllLossBackward0>) 95400 tensor(4.7030, device='cuda:0', grad_fn=<NllLossBackward0>) 95500 tensor(4.9484, device='cuda:0', grad_fn=<NllLossBackward0>) 342000 95600 tensor(4.9381, device='cuda:0', grad_fn=<NllLossBackward0>) 95700 tensor(5.1246, device='cuda:0', grad_fn=<NllLossBackward0>) 95800 tensor(5.0476, device='cuda:0', grad_fn=<NllLossBackward0>) 343000 95900 tensor(4.8117, device='cuda:0', grad_fn=<NllLossBackward0>) 96000 tensor(5.1602, device='cuda:0', grad_fn=<NllLossBackward0>) 344000 96100 tensor(4.9363, device='cuda:0', grad_fn=<NllLossBackward0>) 96200 tensor(4.9611, device='cuda:0', grad_fn=<NllLossBackward0>) 96300 tensor(5.0053, device='cuda:0', grad_fn=<NllLossBackward0>) 345000 96400 tensor(5.0561, device='cuda:0', grad_fn=<NllLossBackward0>) 96500 tensor(5.2501, device='cuda:0', grad_fn=<NllLossBackward0>) 96600 tensor(5.0376, device='cuda:0', grad_fn=<NllLossBackward0>) 346000 96700 tensor(5.1798, device='cuda:0', grad_fn=<NllLossBackward0>) 96800 tensor(5.0684, device='cuda:0', grad_fn=<NllLossBackward0>) 96900 tensor(5.2261, device='cuda:0', grad_fn=<NllLossBackward0>) 347000 97000 tensor(4.9142, device='cuda:0', grad_fn=<NllLossBackward0>) 97100 tensor(5.0132, device='cuda:0', grad_fn=<NllLossBackward0>) 97200 tensor(5.1487, device='cuda:0', grad_fn=<NllLossBackward0>) 348000 97300 tensor(4.9732, device='cuda:0', grad_fn=<NllLossBackward0>) 97400 tensor(5.0108, device='cuda:0', grad_fn=<NllLossBackward0>) 349000 97500 tensor(5.1571, device='cuda:0', grad_fn=<NllLossBackward0>) 97600 tensor(4.9469, device='cuda:0', grad_fn=<NllLossBackward0>) 97700 tensor(5.0162, device='cuda:0', grad_fn=<NllLossBackward0>) 350000 97800 tensor(4.9770, device='cuda:0', grad_fn=<NllLossBackward0>) 97900 tensor(4.7823, device='cuda:0', grad_fn=<NllLossBackward0>) 98000 tensor(4.8846, device='cuda:0', grad_fn=<NllLossBackward0>) 351000 98100 tensor(5.2296, device='cuda:0', grad_fn=<NllLossBackward0>) 98200 tensor(4.9268, device='cuda:0', grad_fn=<NllLossBackward0>) 98300 tensor(5.0850, device='cuda:0', grad_fn=<NllLossBackward0>) 352000 98400 tensor(5.1678, device='cuda:0', grad_fn=<NllLossBackward0>) 98500 tensor(5.2236, device='cuda:0', grad_fn=<NllLossBackward0>) 353000 98600 tensor(5.2631, device='cuda:0', grad_fn=<NllLossBackward0>) 98700 tensor(5.0668, device='cuda:0', grad_fn=<NllLossBackward0>) 98800 tensor(5.0674, device='cuda:0', grad_fn=<NllLossBackward0>) 354000 98900 tensor(5.0695, device='cuda:0', grad_fn=<NllLossBackward0>) 99000 tensor(5.2623, device='cuda:0', grad_fn=<NllLossBackward0>) 99100 tensor(5.1354, device='cuda:0', grad_fn=<NllLossBackward0>) 355000 99200 tensor(5.1557, device='cuda:0', grad_fn=<NllLossBackward0>) 99300 tensor(4.9458, device='cuda:0', grad_fn=<NllLossBackward0>) 99400 tensor(5.2565, device='cuda:0', grad_fn=<NllLossBackward0>) 356000 99500 tensor(5.2732, device='cuda:0', grad_fn=<NllLossBackward0>) 99600 tensor(5.0011, device='cuda:0', grad_fn=<NllLossBackward0>) 99700 tensor(4.8497, device='cuda:0', grad_fn=<NllLossBackward0>) 357000 99800 tensor(5.0679, device='cuda:0', grad_fn=<NllLossBackward0>) 99900 tensor(5.0005, device='cuda:0', grad_fn=<NllLossBackward0>) 100000 tensor(4.9254, device='cuda:0', grad_fn=<NllLossBackward0>) 358000 100100 tensor(5.4405, device='cuda:0', grad_fn=<NllLossBackward0>) 100200 tensor(4.7585, device='cuda:0', grad_fn=<NllLossBackward0>) 359000 100300 tensor(5.0083, device='cuda:0', grad_fn=<NllLossBackward0>) 100400 tensor(4.9893, device='cuda:0', grad_fn=<NllLossBackward0>) 100500 tensor(4.8576, device='cuda:0', grad_fn=<NllLossBackward0>) 360000 100600 tensor(4.8928, device='cuda:0', grad_fn=<NllLossBackward0>) 100700 tensor(4.8947, device='cuda:0', grad_fn=<NllLossBackward0>) 100800 tensor(5.2061, device='cuda:0', grad_fn=<NllLossBackward0>) 361000 100900 tensor(5.1799, device='cuda:0', grad_fn=<NllLossBackward0>) 101000 tensor(5.0292, device='cuda:0', grad_fn=<NllLossBackward0>) 101100 tensor(5.0604, device='cuda:0', grad_fn=<NllLossBackward0>) 362000 101200 tensor(5.0287, device='cuda:0', grad_fn=<NllLossBackward0>) 101300 tensor(5.2627, device='cuda:0', grad_fn=<NllLossBackward0>) 363000 101400 tensor(4.8027, device='cuda:0', grad_fn=<NllLossBackward0>) 101500 tensor(5.0308, device='cuda:0', grad_fn=<NllLossBackward0>) 101600 tensor(5.2625, device='cuda:0', grad_fn=<NllLossBackward0>) 364000 101700 tensor(4.9086, device='cuda:0', grad_fn=<NllLossBackward0>) 101800 tensor(4.7064, device='cuda:0', grad_fn=<NllLossBackward0>) 101900 tensor(4.9304, device='cuda:0', grad_fn=<NllLossBackward0>) 365000 102000 tensor(5.1991, device='cuda:0', grad_fn=<NllLossBackward0>) 102100 tensor(5.1194, device='cuda:0', grad_fn=<NllLossBackward0>) 102200 tensor(5.3738, device='cuda:0', grad_fn=<NllLossBackward0>) 366000 102300 tensor(5.2404, device='cuda:0', grad_fn=<NllLossBackward0>) 102400 tensor(5.1821, device='cuda:0', grad_fn=<NllLossBackward0>) 102500 tensor(5.1676, device='cuda:0', grad_fn=<NllLossBackward0>) 367000 102600 tensor(5.1029, device='cuda:0', grad_fn=<NllLossBackward0>) 102700 tensor(5.1796, device='cuda:0', grad_fn=<NllLossBackward0>) 368000 102800 tensor(5.0834, device='cuda:0', grad_fn=<NllLossBackward0>) 102900 tensor(4.8583, device='cuda:0', grad_fn=<NllLossBackward0>) 103000 tensor(5.0342, device='cuda:0', grad_fn=<NllLossBackward0>) 369000 103100 tensor(5.3343, device='cuda:0', grad_fn=<NllLossBackward0>) 103200 tensor(4.9536, device='cuda:0', grad_fn=<NllLossBackward0>) 103300 tensor(5.0340, device='cuda:0', grad_fn=<NllLossBackward0>) 370000 103400 tensor(5.0580, device='cuda:0', grad_fn=<NllLossBackward0>) 103500 tensor(4.9467, device='cuda:0', grad_fn=<NllLossBackward0>) 103600 tensor(5.0845, device='cuda:0', grad_fn=<NllLossBackward0>) 371000 103700 tensor(5.1711, device='cuda:0', grad_fn=<NllLossBackward0>) 103800 tensor(5.0389, device='cuda:0', grad_fn=<NllLossBackward0>) 103900 tensor(4.9456, device='cuda:0', grad_fn=<NllLossBackward0>) 372000 104000 tensor(4.9239, device='cuda:0', grad_fn=<NllLossBackward0>) 104100 tensor(4.9678, device='cuda:0', grad_fn=<NllLossBackward0>) 373000 104200 tensor(5.0781, device='cuda:0', grad_fn=<NllLossBackward0>) 104300 tensor(4.8800, device='cuda:0', grad_fn=<NllLossBackward0>) 104400 tensor(5.2081, device='cuda:0', grad_fn=<NllLossBackward0>) 374000 104500 tensor(5.1255, device='cuda:0', grad_fn=<NllLossBackward0>) 104600 tensor(4.9805, device='cuda:0', grad_fn=<NllLossBackward0>) 104700 tensor(4.9825, device='cuda:0', grad_fn=<NllLossBackward0>) 375000 104800 tensor(5.5062, device='cuda:0', grad_fn=<NllLossBackward0>) 104900 tensor(5.0209, device='cuda:0', grad_fn=<NllLossBackward0>) 105000 tensor(5.1373, device='cuda:0', grad_fn=<NllLossBackward0>) 376000 105100 tensor(5.0935, device='cuda:0', grad_fn=<NllLossBackward0>) 105200 tensor(5.1819, device='cuda:0', grad_fn=<NllLossBackward0>) 105300 tensor(5.1386, device='cuda:0', grad_fn=<NllLossBackward0>) 377000 105400 tensor(5.3493, device='cuda:0', grad_fn=<NllLossBackward0>) 105500 tensor(5.2102, device='cuda:0', grad_fn=<NllLossBackward0>) 378000 105600 tensor(4.9484, device='cuda:0', grad_fn=<NllLossBackward0>) 105700 tensor(4.9119, device='cuda:0', grad_fn=<NllLossBackward0>) 105800 tensor(4.9584, device='cuda:0', grad_fn=<NllLossBackward0>) 379000 105900 tensor(5.1010, device='cuda:0', grad_fn=<NllLossBackward0>) 106000 tensor(5.1442, device='cuda:0', grad_fn=<NllLossBackward0>) 106100 tensor(5.3454, device='cuda:0', grad_fn=<NllLossBackward0>) 380000 106200 tensor(5.0330, device='cuda:0', grad_fn=<NllLossBackward0>) 106300 tensor(5.1396, device='cuda:0', grad_fn=<NllLossBackward0>) 106400 tensor(4.9693, device='cuda:0', grad_fn=<NllLossBackward0>) 381000 106500 tensor(5.2242, device='cuda:0', grad_fn=<NllLossBackward0>) 106600 tensor(4.9223, device='cuda:0', grad_fn=<NllLossBackward0>) 106700 tensor(5.1467, device='cuda:0', grad_fn=<NllLossBackward0>) 382000 106800 tensor(4.9459, device='cuda:0', grad_fn=<NllLossBackward0>) 106900 tensor(5.2412, device='cuda:0', grad_fn=<NllLossBackward0>) 383000 107000 tensor(4.9030, device='cuda:0', grad_fn=<NllLossBackward0>) 107100 tensor(5.1487, device='cuda:0', grad_fn=<NllLossBackward0>) 107200 tensor(5.1303, device='cuda:0', grad_fn=<NllLossBackward0>) 384000 107300 tensor(4.8918, device='cuda:0', grad_fn=<NllLossBackward0>) 107400 tensor(5.0686, device='cuda:0', grad_fn=<NllLossBackward0>) 107500 tensor(4.8140, device='cuda:0', grad_fn=<NllLossBackward0>) 385000 107600 tensor(4.8693, device='cuda:0', grad_fn=<NllLossBackward0>) 107700 tensor(5.0879, device='cuda:0', grad_fn=<NllLossBackward0>) 107800 tensor(5.0939, device='cuda:0', grad_fn=<NllLossBackward0>) 386000 107900 tensor(5.3087, device='cuda:0', grad_fn=<NllLossBackward0>) 108000 tensor(5.0659, device='cuda:0', grad_fn=<NllLossBackward0>) 108100 tensor(5.0273, device='cuda:0', grad_fn=<NllLossBackward0>) 387000 108200 tensor(5.0357, device='cuda:0', grad_fn=<NllLossBackward0>) 108300 tensor(5.0666, device='cuda:0', grad_fn=<NllLossBackward0>) 388000 108400 tensor(5.1699, device='cuda:0', grad_fn=<NllLossBackward0>) 108500 tensor(4.8493, device='cuda:0', grad_fn=<NllLossBackward0>) 108600 tensor(5.0399, device='cuda:0', grad_fn=<NllLossBackward0>) 389000 108700 tensor(4.5729, device='cuda:0', grad_fn=<NllLossBackward0>) 108800 tensor(4.9686, device='cuda:0', grad_fn=<NllLossBackward0>) 108900 tensor(4.7163, device='cuda:0', grad_fn=<NllLossBackward0>) 390000 109000 tensor(5.1307, device='cuda:0', grad_fn=<NllLossBackward0>) 109100 tensor(4.7140, device='cuda:0', grad_fn=<NllLossBackward0>) 109200 tensor(4.9282, device='cuda:0', grad_fn=<NllLossBackward0>) 391000 109300 tensor(4.5945, device='cuda:0', grad_fn=<NllLossBackward0>) 109400 tensor(4.9045, device='cuda:0', grad_fn=<NllLossBackward0>) 109500 tensor(4.8345, device='cuda:0', grad_fn=<NllLossBackward0>) 392000 109600 tensor(5.0347, device='cuda:0', grad_fn=<NllLossBackward0>) 109700 tensor(5.3972, device='cuda:0', grad_fn=<NllLossBackward0>) 393000 109800 tensor(4.7386, device='cuda:0', grad_fn=<NllLossBackward0>) 109900 tensor(4.9312, device='cuda:0', grad_fn=<NllLossBackward0>) 110000 tensor(5.0107, device='cuda:0', grad_fn=<NllLossBackward0>) 394000 110100 tensor(4.9484, device='cuda:0', grad_fn=<NllLossBackward0>) 110200 tensor(5.3088, device='cuda:0', grad_fn=<NllLossBackward0>) 110300 tensor(5.1758, device='cuda:0', grad_fn=<NllLossBackward0>) 395000 110400 tensor(4.6608, device='cuda:0', grad_fn=<NllLossBackward0>) 110500 tensor(5.0811, device='cuda:0', grad_fn=<NllLossBackward0>) 110600 tensor(4.9037, device='cuda:0', grad_fn=<NllLossBackward0>) 396000 110700 tensor(4.8701, device='cuda:0', grad_fn=<NllLossBackward0>) 110800 tensor(5.3046, device='cuda:0', grad_fn=<NllLossBackward0>) 110900 tensor(4.9689, device='cuda:0', grad_fn=<NllLossBackward0>) 397000 111000 tensor(5.1637, device='cuda:0', grad_fn=<NllLossBackward0>) 111100 tensor(4.9131, device='cuda:0', grad_fn=<NllLossBackward0>) 398000 111200 tensor(4.9308, device='cuda:0', grad_fn=<NllLossBackward0>) 111300 tensor(5.0290, device='cuda:0', grad_fn=<NllLossBackward0>) 111400 tensor(5.1249, device='cuda:0', grad_fn=<NllLossBackward0>) 399000 111500 tensor(5.2130, device='cuda:0', grad_fn=<NllLossBackward0>) 111600 tensor(5.2382, device='cuda:0', grad_fn=<NllLossBackward0>) 111700 tensor(4.8168, device='cuda:0', grad_fn=<NllLossBackward0>) 400000 111800 tensor(5.0276, device='cuda:0', grad_fn=<NllLossBackward0>) 111900 tensor(5.1121, device='cuda:0', grad_fn=<NllLossBackward0>) 112000 tensor(5.0463, device='cuda:0', grad_fn=<NllLossBackward0>) 401000 112100 tensor(4.6612, device='cuda:0', grad_fn=<NllLossBackward0>) 112200 tensor(5.0632, device='cuda:0', grad_fn=<NllLossBackward0>) 402000 112300 tensor(5.2173, device='cuda:0', grad_fn=<NllLossBackward0>) 112400 tensor(5.2271, device='cuda:0', grad_fn=<NllLossBackward0>) 112500 tensor(4.8618, device='cuda:0', grad_fn=<NllLossBackward0>) 403000 112600 tensor(4.9399, device='cuda:0', grad_fn=<NllLossBackward0>) 112700 tensor(4.9514, device='cuda:0', grad_fn=<NllLossBackward0>) 112800 tensor(4.5700, device='cuda:0', grad_fn=<NllLossBackward0>) 404000 112900 tensor(4.8508, device='cuda:0', grad_fn=<NllLossBackward0>) 113000 tensor(5.0784, device='cuda:0', grad_fn=<NllLossBackward0>) 113100 tensor(5.0205, device='cuda:0', grad_fn=<NllLossBackward0>) 405000 113200 tensor(4.5998, device='cuda:0', grad_fn=<NllLossBackward0>) 113300 tensor(4.9371, device='cuda:0', grad_fn=<NllLossBackward0>) 113400 tensor(5.0241, device='cuda:0', grad_fn=<NllLossBackward0>) 406000 113500 tensor(5.2931, device='cuda:0', grad_fn=<NllLossBackward0>) 113600 tensor(5.2488, device='cuda:0', grad_fn=<NllLossBackward0>) 407000 113700 tensor(4.8483, device='cuda:0', grad_fn=<NllLossBackward0>) 113800 tensor(5.2471, device='cuda:0', grad_fn=<NllLossBackward0>) 113900 tensor(5.0391, device='cuda:0', grad_fn=<NllLossBackward0>) 408000 114000 tensor(4.6978, device='cuda:0', grad_fn=<NllLossBackward0>) 114100 tensor(4.8170, device='cuda:0', grad_fn=<NllLossBackward0>) 114200 tensor(4.7943, device='cuda:0', grad_fn=<NllLossBackward0>) 409000 114300 tensor(4.6678, device='cuda:0', grad_fn=<NllLossBackward0>) 114400 tensor(4.8687, device='cuda:0', grad_fn=<NllLossBackward0>) 114500 tensor(4.9471, device='cuda:0', grad_fn=<NllLossBackward0>) 410000 114600 tensor(5.1608, device='cuda:0', grad_fn=<NllLossBackward0>) 114700 tensor(5.1240, device='cuda:0', grad_fn=<NllLossBackward0>) 114800 tensor(5.0586, device='cuda:0', grad_fn=<NllLossBackward0>) 411000 114900 tensor(5.0095, device='cuda:0', grad_fn=<NllLossBackward0>) 115000 tensor(4.9676, device='cuda:0', grad_fn=<NllLossBackward0>) 412000 115100 tensor(5.0771, device='cuda:0', grad_fn=<NllLossBackward0>) 115200 tensor(5.0396, device='cuda:0', grad_fn=<NllLossBackward0>) 115300 tensor(5.0663, device='cuda:0', grad_fn=<NllLossBackward0>) 413000 115400 tensor(5.2931, device='cuda:0', grad_fn=<NllLossBackward0>) 115500 tensor(5.2901, device='cuda:0', grad_fn=<NllLossBackward0>) 115600 tensor(5.0953, device='cuda:0', grad_fn=<NllLossBackward0>) 414000 115700 tensor(4.9706, device='cuda:0', grad_fn=<NllLossBackward0>) 115800 tensor(5.0768, device='cuda:0', grad_fn=<NllLossBackward0>) 115900 tensor(5.1668, device='cuda:0', grad_fn=<NllLossBackward0>) 415000 116000 tensor(5.0913, device='cuda:0', grad_fn=<NllLossBackward0>) 116100 tensor(4.8294, device='cuda:0', grad_fn=<NllLossBackward0>) 116200 tensor(5.2189, device='cuda:0', grad_fn=<NllLossBackward0>) 416000 116300 tensor(4.3633, device='cuda:0', grad_fn=<NllLossBackward0>) 116400 tensor(5.1168, device='cuda:0', grad_fn=<NllLossBackward0>) 417000 116500 tensor(5.0554, device='cuda:0', grad_fn=<NllLossBackward0>) 116600 tensor(5.1477, device='cuda:0', grad_fn=<NllLossBackward0>) 116700 tensor(4.9884, device='cuda:0', grad_fn=<NllLossBackward0>) 418000 116800 tensor(4.8381, device='cuda:0', grad_fn=<NllLossBackward0>) 116900 tensor(5.0709, device='cuda:0', grad_fn=<NllLossBackward0>) 117000 tensor(5.0518, device='cuda:0', grad_fn=<NllLossBackward0>) 419000 117100 tensor(5.0967, device='cuda:0', grad_fn=<NllLossBackward0>) 117200 tensor(4.9430, device='cuda:0', grad_fn=<NllLossBackward0>) 117300 tensor(5.2903, device='cuda:0', grad_fn=<NllLossBackward0>) 420000 117400 tensor(5.2733, device='cuda:0', grad_fn=<NllLossBackward0>) 117500 tensor(4.8548, device='cuda:0', grad_fn=<NllLossBackward0>) 421000 117600 tensor(5.1768, device='cuda:0', grad_fn=<NllLossBackward0>) 117700 tensor(5.1013, device='cuda:0', grad_fn=<NllLossBackward0>) 117800 tensor(4.9804, device='cuda:0', grad_fn=<NllLossBackward0>) 422000 117900 tensor(5.0671, device='cuda:0', grad_fn=<NllLossBackward0>) 118000 tensor(5.1332, device='cuda:0', grad_fn=<NllLossBackward0>) 118100 tensor(4.9436, device='cuda:0', grad_fn=<NllLossBackward0>) 423000 118200 tensor(4.9859, device='cuda:0', grad_fn=<NllLossBackward0>) 118300 tensor(4.9755, device='cuda:0', grad_fn=<NllLossBackward0>) 118400 tensor(5.0902, device='cuda:0', grad_fn=<NllLossBackward0>) 424000 118500 tensor(4.9845, device='cuda:0', grad_fn=<NllLossBackward0>) 118600 tensor(5.1098, device='cuda:0', grad_fn=<NllLossBackward0>) 118700 tensor(5.1192, device='cuda:0', grad_fn=<NllLossBackward0>) 425000 118800 tensor(5.1445, device='cuda:0', grad_fn=<NllLossBackward0>) 118900 tensor(5.1731, device='cuda:0', grad_fn=<NllLossBackward0>) 426000 119000 tensor(4.6613, device='cuda:0', grad_fn=<NllLossBackward0>) 119100 tensor(4.9640, device='cuda:0', grad_fn=<NllLossBackward0>) 119200 tensor(5.2998, device='cuda:0', grad_fn=<NllLossBackward0>) 427000 119300 tensor(4.9427, device='cuda:0', grad_fn=<NllLossBackward0>) 119400 tensor(4.9993, device='cuda:0', grad_fn=<NllLossBackward0>) 119500 tensor(5.0117, device='cuda:0', grad_fn=<NllLossBackward0>) 428000 119600 tensor(5.3324, device='cuda:0', grad_fn=<NllLossBackward0>) 119700 tensor(4.9375, device='cuda:0', grad_fn=<NllLossBackward0>) 119800 tensor(5.0221, device='cuda:0', grad_fn=<NllLossBackward0>) 429000 119900 tensor(4.7975, device='cuda:0', grad_fn=<NllLossBackward0>) 120000 tensor(4.9253, device='cuda:0', grad_fn=<NllLossBackward0>) 120100 tensor(4.9377, device='cuda:0', grad_fn=<NllLossBackward0>) 430000 120200 tensor(5.0909, device='cuda:0', grad_fn=<NllLossBackward0>) 120300 tensor(4.9946, device='cuda:0', grad_fn=<NllLossBackward0>) 431000 120400 tensor(5.3612, device='cuda:0', grad_fn=<NllLossBackward0>) 120500 tensor(4.9678, device='cuda:0', grad_fn=<NllLossBackward0>) 120600 tensor(5.2217, device='cuda:0', grad_fn=<NllLossBackward0>) 432000
Eval
model = Bigram(vocab_size, embed_size).to(device)
model.load_state_dict(torch.load('model.bin'))
model.eval()
res = torch.tensor(vocab.forward(['for'])).to(device)
out = model(res)
top = torch.topk(out[0], 10)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))/usr/local/lib/python3.10/dist-packages/torch/nn/modules/container.py:217: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument. input = module(input)
[('<unk>', 0, 0.23156249523162842),
('the', 1, 0.2045561522245407),
('a', 5, 0.0636623203754425),
('his', 20, 0.012841351330280304),
('their', 40, 0.012044394388794899),
('this', 28, 0.011758995242416859),
('tho', 33, 0.010536346584558487),
('some', 77, 0.008259670808911324),
('any', 49, 0.007337945979088545),
('an', 38, 0.007214350625872612)]vocab = train_dataset.vocab
res = torch.tensor(vocab.forward(['wait'])).to(device)
out = model(res)
top = torch.topk(out[0], 20)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))[('until', 145, 0.17603398859500885),
('for', 9, 0.16400693356990814),
('<unk>', 0, 0.15191353857517242),
('on', 15, 0.04624223709106445),
('till', 677, 0.035729214549064636),
('a', 5, 0.03367603197693825),
('to', 4, 0.029361305758357048),
('upon', 59, 0.01995147578418255),
('and', 3, 0.01906605064868927),
('in', 6, 0.013167516328394413),
('at', 14, 0.011669990606606007),
('the', 1, 0.010971800424158573),
('of', 2, 0.005925077944993973),
('with', 16, 0.0055325529538095),
('In', 32, 0.004919056314975023),
('until\\\\nthe', 5509, 0.004719363059848547),
('tor', 532, 0.004647853318601847),
('for\\\\nthe', 389, 0.004400868900120258),
('two', 74, 0.0043497709557414055),
('patiently', 14401, 0.004239553119987249)]cos = nn.CosineSimilarity(dim=1, eps=1e-6)
embeddings = model.model[0].weight
vec = embeddings[vocab['take']]
similarities = cos(vec, embeddings)
top = torch.topk(similarities, 10)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))[('take', 152, 1.0000001192092896),
('took', 248, 0.8024641275405884),
('taking', 538, 0.775847852230072),
('takes', 1108, 0.7307196855545044),
('to\\\\ntake', 3165, 0.7301462888717651),
('taken', 180, 0.6280043125152588),
('will\\\\ntake', 11101, 0.6152595281600952),
('tako', 6281, 0.5979241132736206),
('have\\\\ntaken', 15483, 0.5244049429893494),
('Take', 5203, 0.5183135867118835)]vocab = train_dataset.vocab
res = torch.tensor(vocab.forward(['take'])).to(device)
out = model(res)
top = torch.topk(out[0], 20)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
list(zip(top_words, top_indices, top_probs))[('<unk>', 0, 0.18006379902362823),
('the', 1, 0.11970410495996475),
('a', 5, 0.07213426381349564),
('care', 611, 0.027887443080544472),
('up', 69, 0.027840441092848778),
('it', 17, 0.023985734209418297),
('place', 159, 0.020694952458143234),
('advantage', 1458, 0.015635941177606583),
('his', 20, 0.014868981204926968),
('part', 131, 0.013506578281521797),
('an', 38, 0.013118326663970947),
('their', 40, 0.010849231854081154),
('hold', 478, 0.010717789642512798),
('them', 72, 0.010186631232500076),
('to', 4, 0.009746687486767769),
('this', 28, 0.009519988670945168),
('any', 49, 0.009436620399355888),
('her', 53, 0.008774512447416782),
('him', 70, 0.008407332003116608),
('all', 34, 0.007673078216612339)]Create files for geval
def get_values(presc_word, model, vocab):
ixs = torch.tensor(vocab.forward([presc_word])).to(device)
out = model(ixs)
top = torch.topk(out[0], 20)
top_indices = top.indices.tolist()
top_probs = top.values.tolist()
top_words = vocab.lookup_tokens(top_indices)
return list(zip(top_words, top_probs))
def last_word(text):
"""Return the last word of a string."""
last_word = ""
for i in range(len(text)-1, -1, -1):
if text[i] == ' ':
return last_word[::-1]
else:
last_word += text[i]
return last_word[::-1]
def first_word(text):
"""Return the first word of a string."""
word = ""
for i in range(len(text)-1):
if text[i] == ' ':
return word
else:
word += text[i]
return word
def sum_prob(dic):
probsum = sum(float(val) for key, val in dic.items())
probsum = probsum - float(dic.get('<unk>', 0))
if "<unk>" in dic.keys():
del dic['<unk>']
tab = [(key, val) for key, val in dic.items()]
tab.append(('<unk>', 1-probsum))
return tab
def format_to(dic):
tab = sum_prob(dic)
result = ''
for element in tab[:-1]:
result+=str(element[0])+':'+str(element[1])+'\t'
result+=':'+ str(tab[-1][1])+'\n'
return resultwith lzma.open(test_file, 'rt') as file:
predict_words = []
results = []
for line in file:
split = line.split('\t')[6:]
predict_words.append(last_word(split[0]))
vocab = train_dataset.vocab
for presc_word in predict_words:
results.append(dict(get_values(presc_word, model ,vocab)))
with open(out_file, 'w') as outfile:
for elem in results:
outfile.write(format_to(elem))