Skip to content

Question about the implementation of mpu.cross_entropy when using tensor parallel #394

Open
@robin087

Description

@robin087

Hello.
When using tensor parallel on bloom (tp_size = 8), we find that the cross_entropy loss computed by mpu.cross_entropy is different from torch.nn.functional.cross_entropy. The difference is about 1% for our data.
For the implementation of mpu.cross_entropy, we find that the loss is computed on the partition_vocab_size which is 8 times smaller than vocab_size (tp_size = 8). We think maybe this implementation causes the difference above.
In this case, is this implementation correct? Or can this implementation ensure the performance when using tensor parallel?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions