How to get the likelihood from the model?

[jasonppy]

Hi authors,

I'm trying to get the likelihood of Pengi on (audio, question, answer) tuples, but wasn't able to do so. Is it possible to have get some help on this?

I think probably this forward function calculated the loss: https://github.com/microsoft/Pengi/blob/main/models/pengi.py#L174
where audio should be the output of preprocess_audio, texts_enc should be the output of running preprocess_text on question, texts_dec should be the output of running preprocess_text on answer. However I wasn't able to loss from the output, even if I pass label = texts_dec['input_ids'] (https://github.com/microsoft/Pengi/blob/main/models/decoder.py#L219) , I still get bugs on dimension when calculating cross_entropy loss

Your help is greatly appreciated.

Best,
Puyuan

[soham97]

Hi @jasonppy, the loss computation during training looks like:

1. outputs = model(audios, texts_enc, texts_dec)
   where model is the PENGI model, audios is float32 tensor, texts_enc and texts_dec are tokenized text input and text output.
2. logits = outputs.logits[:, total_prefix_length - 1: -1]
   Remove the outputs corresponding to the total prefix length. This is equal to length of audio projection and length of input text
3. loss = F.cross_entropy(logits.reshape(-1, logits.shape[-1]), texts_dec['input_ids'].flatten(), ignore_index=0)
   Compute cross entropy per token and average. Swap 0 with whichever token index is used for padding.

For texts_dec in step 1, make sure to add/prepend ones equal to the total prefix length to the attention mask of tokenized text.