Contrastive-Predictive-Coding-PyTorch

Contrastive-Predictive-Coding-PyTorch copied to clipboard

In the implementation, a random position of the sequence is used to compute the NCE loss. The paper mentions that the GRU output at each step is used to predict...

Ieremie

The paper does not mention the use of Batch Normalization in the case of the audio task. In the case of the Vision task, it mentions that '' We did...

Ieremie

In the calculation of the NCE loss, the softmax does not have a dimension to compute the result and by default, PyTorch uses dim=1 with 2D input. The Loss in...

Ieremie

Feed entire input to encoder??

1

I see in your implementation that you feed entire signal into the encoder, while the paper has noted that each timestemp should be insert seperatly. When you feed the entire...

NeteeraAF

Train and test data not available

1

Dear Jeff, Thank you so much for providing this great repository! Sincerely appreciate your great implementation! However, after reading all the closed issues and trying out for initializing the training,...

martinmamql

What is the format of "list" file?

2

I saw list files such as "LibriSpeech/list/train.txt" are required parameters for `main.py`. It seems such files are not provided by librispeech officially. What is the format of them? Could you...

couragelfyang

At [Line 310](https://github.com/jefflai108/Contrastive-Predictive-Coding-PyTorch/blob/a9dab4e759aaa68dce1b1ada46a8035076ba3296/src/model/model.py#L310), you have the following code ```python output, hidden = self.gru(forward_seq, hidden) # output size e.g. 8*100*256 c_t = output[:,t_samples,:].view(batch, 256) # c_t e.g. size 8*256 ``` So...

KinWaiCheuk

Hey should 'correct' only be if the final prediction was correct? Same for accuracy?

1

https://github.com/jefflai108/Contrastive-Predictive-Coding-PyTorch/blob/a9dab4e759aaa68dce1b1ada46a8035076ba3296/src/model/model.py#L100 Meaning, should that `correct` variable be += ?

colinator

NCE accuracy calculation

1

Hello @jefflai108, I think the accuracy also should have its denominator as batch*self.timestep at https://github.com/jefflai108/Contrastive-Predictive-Coding-PyTorch/blob/a9dab4e759aaa68dce1b1ada46a8035076ba3296/src/model/model.py#L320 Possibly for other models too although I did not check them.

JaejinCho

the self.softmax() don't update the paras in the CDCK2?

1

I don't understand how to update the self.softmax() when training, because the acc don't backward. How does it work?

super-wcg