Skip to content

when will AVX512_BF16 be supported? #9

Description

@marsggbo

Hi authors,

Thanks for sharing this codebase! In your paper you mention:

"we design a specialized computation kernel for expert processing on the CPU using the AVX512_BF16 instruction set, which is not supported in the native PyTorch implementation."

However, in the released repository I only see the following Python function:

def run_expert_at_cpu(self, i_layer, i_expert, inps, routing_weights):
    """Run the expert at CPU"""
    return self.model.layers[i_layer].block_sparse_moe.experts[i_expert](
        inps, routing_weights
    )

May I ask when will AVX512_BF16 be supported?

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

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions