barissayil / sentimentanalysis Goto Github PK

My (limited) understanding is that in the transfomers code some pretrained models have renamed or moved. huggingface/transformers#4636

Can you achieve three classifications?positive、negative and neutral

Hi,

Would it be possible for you to share pre-trained model weights for this model?

Thanks

Hi @barissayil,

Thanks you for sharing your excellent work. I met an error of "string indices must be integers" When I run the analyze.py file. The error is showed below:

The error are list below:
Please wait while the analyser is being prepared.
Input sentiment to analyze: I am happy.
Traceback (most recent call last):
 File "analyze.py", line 53, in <module>
   classify_sentiment(sentence)
 File "analyze.py", line 14, in classify_sentiment
   logit = model(input_ids=input_ids, attention_mask=attention_mask)
 File "/Users/insight/opt/anaconda3/envs/meetingminutes-sentiment/lib/python3.6/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
   result = self.forward(*input, **kwargs)
 File "/Users/insight/Projects/MeetingMinutes-Minning/bert-extractive-summarizer/sentiment/SentimentAnalysis/modeling.py", line 23, in forward
   cls_reps = reps[:, 0]
TypeError: string indices must be integers

Could you please give me some suggestion to find the solution? Thanks a lot!

Hey, thanks for the code.

When I was running, there was some error. Could you please help me to check it. Thank

File "F:\PycharmProjects\SentimentAnalysis-master\modeling.py", line 25, in forward cls_reps = outputs.last_hidden_state[:, 0] AttributeError: 'tuple' object has no attribute 'last_hidden_state'

Hey barissayil,

Can you please include in the repository a setup.py file ? I cant use pip to install it as a package.

Thanks,
Roi.

Can I use this model for direct inference on japanese text or is re-training necessary ?

Hi, when I uncommented the choosing device part and tried to execute the analize.py, I received runtime error, it seems some of the tensors are still on cpu.
# device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# model = model.to(device)
"
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument index in method wrapper__index_select)
"
So do you know how to solve this?

When I run train.py the following error occurs:

Epoch 0 complete! Validation Accuracy : 0.8917410969734192, Validation Loss : 0.27176217176020145
Best validation accuracy improved from 0 to 0.8917410969734192, saving model...
Traceback (most recent call last):
File "train.py", line 106, in
train(net, criterion, optimizer, train_loader, val_loader, args)
File "train.py", line 77, in train
torch.save(net.state_dict(), 'models/model'.format(ep, args.freeze_bert))
NameError: name 'ep' is not defined