search

microsoft / torchscale

Foundation Architecture for (M)LLMs
https://aka.ms/GeneralAI
MIT License
3k stars 201 forks source link

EncoderDecoder Configuration Issue #32

Closed klae01 closed 1 year ago

klae01 commented 1 year ago

I've identified a problem in the EncoderDecoderConfig class within the architecture module of the torchscale package.

The EncoderDecoderConfig class currently does not contain the normalize_output element. This missing element is causing some functionality of the package to not work as expected. Specifically, when the EncoderDecoder class is used with a EncoderDecoderConfig object, the lack of the normalize_output element can lead to incorrect behavior.

I recommend adding the normalize_output element to the EncoderDecoderConfig class. I believe that this change will resolve the issue and make the EncoderDecoder class function as expected.

Furthermore, I have added a new GitHub Action in pull request #31 to help prevent issues like this in the future.

shumingma commented 1 year ago

Thank you for the contribution! Already fixed the config issue and merged the PR.