iliazenkov / transformer-cnn-emotion-recognition Goto Github PK

How do i get this problem?? I want to solve the problem

I'm getting the following error in the Load the Trained Model from Checkpoint for Evaluation section when trying to load the pre-trained models in the repo.

# pick load folder  
load_folder = '/content/gdrive/My Drive/DL/models/checkpoints'  

# pick the epoch to load
epoch = '429'
model_name = f'parallel_all_you_wantFINAL-{epoch}.pkl'

# make full load path
load_path = os.path.join(load_folder, model_name)

## instantiate empty model and populate with params from binary 
model = parallel_all_you_want(len(emotions_dict))
load_checkpoint(optimizer, model, load_path)

print(f'Loaded model from {load_path}')

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-51-de60b48d9410> in <module>
     11 ## instantiate empty model and populate with params from binary
     12 model = parallel_all_you_want(len(emotions_dict))
---> 13 load_checkpoint(optimizer, model, load_path)
     14 
     15 print(f'Loaded model from {load_path}')

<ipython-input-50-2469ccaf4901> in load_checkpoint(optimizer, model, filename)
     11 def load_checkpoint(optimizer, model, filename):
     12     checkpoint_dict = torch.load(filename, map_location=torch.device('cpu'))
---> 13     epoch = checkpoint_dict['epoch']
     14     model.load_state_dict(checkpoint_dict['model'])
     15     if optimizer is not None:

KeyError: 'epoch'

It appears that the pre-trained model files were not generated using the structure expected by the notebook.

Hi IliaZenkov：
I wanma quote your papers (traditional method and this method), you can send the DOI of your papers.
Looking forward to your reply

Hi, congrats on the project!
I am getting this error, could not figure out what's going on:

File "/media/user/hk-data/PycharmProjects/dnn_embeddings_pytorch/train_model.py", line 148, in train_model
output = net(x_train)
File "/home/user/anaconda3/envs/general_py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/media/user/hk-data/PycharmProjects/dnn_embeddings_pytorch/dnn_models.py", line 240, in forward
output_logits = self.fc1_linear(complete_embedding)
File "/home/user/anaconda3/envs/general_py37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/user/anaconda3/envs/general_py37/lib/python3.7/site-packages/torch/nn/modules/linear.py", line 93, in forward
return F.linear(input, self.weight, self.bias)
File "/home/user/anaconda3/envs/general_py37/lib/python3.7/site-packages/torch/nn/functional.py", line 1690, in linear
ret = torch.addmm(bias, input, weight.t())
RuntimeError: mat1 dim 1 must match mat2 dim 0

Do you not need to use Position Encoding when using Transformer?

Thanks for the code.

I get different results than yours, even though I use the same dataset.

As you can see, the validation loss slightly increases with epocs, while it decreases in your result.
I didn't change any parameters or anything else in the code, only the directories.