gsp-27 / pytorch_squeezenet Goto Github PK

NameError Traceback (most recent call last)
~\pytorch_Squeezenet-master\main.py in
204 fig2, ax2 = plt.subplots()
205 train_acc, val_acc = list(), list()
--> 206 for i in xrange(1,args.epoch+1):
207 train_acc.append(train(i))
208 val_acc.append(val())

NameError: name 'xrange' is not defined

Sorry to disturb. I tried to run squeezenet using pytorch on GPU. And I got this error at the end of Epoch 1.
Train Epoch: 1 [48992/50000 (98%)] Loss: 1.618558
Train Epoch: 1 [49312/50000 (99%)] Loss: 1.442842
Train Epoch: 1 [49632/50000 (99%)] Loss: 1.601359
Train Epoch: 1 [49952/50000 (100%)] Loss: 1.511631
Traceback (most recent call last):
File "main.py", line 228, in
train_acc.append(train(i))
File "main.py", line 131, in train
scores = scores.view(args.batch_size, args.num_classes)
RuntimeError: shape '[32, 10]' is invalid for input of size 160
Thank you for any suggestion.

Please tell me how to use other data sets, which are csv files

when i use a batch-size of 64, an runtime error occurs when it will almost finish the firt epoch training:
Train Epoch: 1 [49344/50000 (99%)] Loss: 1.619473
Train Epoch: 1 [49984/50000 (100%)] Loss: 1.664685
Traceback (most recent call last):
File "F:/paper_code/pytorch_Squeezenet/main.py", line 208, in
train_acc.append(train(i))
File "F:/paper_code/pytorch_Squeezenet/main.py", line 126, in train
scores = scores.view(args.batch_size, args.num_classes)
RuntimeError: shape '[64, 10]' is invalid for input of size 160

can anyone tell me what i should do?QAQ