Skip to content

Bug in sample_replay function (zero-dimensional tensor concatenation) #2

Description

@Diana-sis

Hi, I'm trying to reproduce HiCL on CIFAR-10. Expert 0 trains successfully (96.4% accuracy), but when starting Expert 1, I get:

RuntimeError: zero-dimensional tensor (at position 0) cannot be concatenated in sample_replay function line 233.

Environment: Python 3.8, PyTorch 2.0.0+cu118, RTX 3090.

Has anyone encountered this? Is there a fix or an updated version of the code?

Thanks!

Activity

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions