Skip to content

【Hackathon 9th No.31】Unit test for gptq_marlin_repack #3

@cloudforge1

Description

@cloudforge1

Add unit tests for the GPTQ Marlin weight repacking op.

Source: custom_ops/gpu_ops/moe/gptq_marlin_repack.cu
Registration: custom_ops/gpu_ops/cpp_extensions.cc
Test file: tests/operators/test_gptq_marlin_repack.py

Should verify that repacking produces the correct weight layout for Marlin GEMM kernels. Test with different quantization group sizes and weight matrix shapes.

Branch: task/031-gptq-marlin-repack-test

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