Skip to content

[BUG] The versions of nvcomp in Docker and the repository are incompatible. #337

@Clebrate

Description

@Clebrate

Describe the bug
When building the project with python setup.py build_ext --inplace inside a Docker container, the compilation fails with a type definition error related to nvcomp library types:
error: ‘nvcompBatchedANSDecompressOpts_t’ does not name a type; did you mean ‘nvcompBatchedANSOpts_t’? error: ‘nvcompBatchedANSCompressOpts_t’ does not name a type; did you mean ‘nvcompBatchedANSOpts_t’?
The build process stops at these errors in the file /recsys-examples/examples/commons/ops/cuda_ops/csrc/kvcache_manager_impl.h, lines 134 and 137.

Steps/Code to reproduce bug

  1. Build the Docker image
    docker build --platform linux/amd64 --build-arg INFERENCEBUILD=1 -t recsys-examples:inference -f docker/Dockerfile .
  2. Run the container and execute the build command
    docker run -it --rm recsys-examples:inference
    cd /recsys-examples/examples/commons
    python setup.py build_ext --inplace
  3. Observe the compilation error shown in the logs.

Expected behavior
The python setup.py build_ext --inplace command should complete successfully, compiling the CUDA extensions without type definition errors related to nvcomp ANS compression/decompression options.

Environment details (please complete the following information):

  • Environment location: Docker
  • Method of recsys-examples install:
    docker build --platform linux/amd64 --build-arg INFERENCEBUILD=1 -t recsys-examples:inference -f docker/Dockerfile .
    @geoffreyQiu

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions