Skip to content

Conversation

@adistomar
Copy link

This PR fixes two things:

  • In generate.py, the state tensor x containing the prompt as well as the masked and unmasked tokens had a hardcoded batch size of 1, meaning that the user cannot perform batching. Fix: set first dim of x to prompt.shape[0] during initialization
  • In eval_llada.py, if the # of processes is 1 (for ex. in single GPU case), then self.accelerator is set to None. In this scenario, performing self.accelerator.wait_for_everyone will cause an error. Fix: add a check for if the variable is None

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant