Commit cc609811 authored by Daniel Fahey's avatar Daniel Fahey
Browse files

python3Packages.vllm: 0.10.1.1 -> 0.10.2 

Notably adds Python 3.13 support (see vllm-project/vllm#13164) among many other things (https://github.com/vllm-project/vllm/releases/tag/v0.10.2

).

Co-authored-by: default avatarGaétan Lepage <gaetan@glepage.com>
parent 94ced647

pkgs/development/python-modules/vllm/default.nix

+18 −12
Original line number Diff line number Diff line
@@ -3,7 +3,7 @@ 
  stdenv,

  python,

  buildPythonPackage,

  pythonAtLeast,

  pythonOlder,

  fetchFromGitHub,

  fetchpatch,

  symlinkJoin,
@@ -115,8 +115,8 @@ let 
    src = fetchFromGitHub {

      owner = "vllm-project";

      repo = "FlashMLA";

      rev = "0e43e774597682284358ff2c54530757b654b8d1";

      hash = "sha256-wxL/jtq/lsLg1o+4392KNgfw5TYlW6lqEVbmR3Jl4/Q=";

      rev = "a757314c04eedd166e329e846c820eb1bdd702de";

      hash = "sha256-KT9R6ju7XzgqKHPGQwzw0yNiKL3DNW6qJrEBvmLn4hY=";

    };



    dontConfigure = true;
@@ -135,15 +135,15 @@ let 
  vllm-flash-attn' = lib.defaultTo (stdenv.mkDerivation {

    pname = "vllm-flash-attn";

    # https://github.com/vllm-project/flash-attention/blob/${src.rev}/vllm_flash_attn/__init__.py

    version = "2.7.4.post1";

    version = "2.7.2.post1";



    # grep for GIT_TAG in the following file

    # https://github.com/vllm-project/vllm/blob/v${version}/cmake/external_projects/vllm_flash_attn.cmake

    src = fetchFromGitHub {

      owner = "vllm-project";

      repo = "flash-attention";

      rev = "57b4e68b9f9d94750b46de8f8dbd2bfcc86edd4f";

      hash = "sha256-c7L7WZVVEnXMOTPBoSp7jhkl9d4TA4sj11QvOSWTDIE=";

      rev = "ee4d25bd84e0cbc7e0b9b9685085fd5db2dcb62a";

      hash = "sha256-2r0Habd/kBpvM4/aQFIYyj+uQAa3M9gjk3DcBZHFNfA=";

    };



    patches = [
@@ -179,7 +179,7 @@ let 


  cpuSupport = !cudaSupport && !rocmSupport;



  # https://github.com/pytorch/pytorch/blob/v2.7.1/torch/utils/cpp_extension.py#L2343-L2345

  # https://github.com/pytorch/pytorch/blob/v2.8.0/torch/utils/cpp_extension.py#L2411-L2414

  supportedTorchCudaCapabilities =

    let

      real = [
@@ -204,8 +204,12 @@ let 
        "10.0a"

        "10.1"

        "10.1a"

        "10.3"

        "10.3a"

        "12.0"

        "12.0a"

        "12.1"

        "12.1a"

      ];

      ptx = lists.map (x: "${x}+PTX") real;

    in
@@ -245,6 +249,7 @@ let 
  mergedCudaLibraries = with cudaPackages; [

    cuda_cudart # cuda_runtime.h, -lcudart

    cuda_cccl

    libcurand # curand_kernel.h

    libcusparse # cusparse.h

    libcusolver # cusolverDn.h

    cuda_nvtx
@@ -266,19 +271,16 @@ in 


buildPythonPackage rec {

  pname = "vllm";

  version = "0.10.1.1";

  version = "0.10.2";

  pyproject = true;



  # https://github.com/vllm-project/vllm/issues/12083

  disabled = pythonAtLeast "3.13";



  stdenv = torch.stdenv;



  src = fetchFromGitHub {

    owner = "vllm-project";

    repo = "vllm";

    tag = "v${version}";

    hash = "sha256-lLNjBv5baER0AArX3IV4HWjDZ2jTGXyGIvnHupR8MGM=";

    hash = "sha256-m9P4cxxdAToGKKIyTQdFupG3vZ3sEueMMxjugYfjKbo=";

  };



  patches = [
@@ -483,6 +485,10 @@ buildPythonPackage rec { 
      lach

      daniel-fahey

    ];

    # Python 3.12 vLLM v0.10.2+CPU blake3 1.0.7 incompatibility

    # discovered during https://github.com/NixOS/nixpkgs/pull/447722

    # reported upstream in https://github.com/vllm-project/vllm/issues/26229

    broken = (cpuSupport && pythonOlder "3.13");

    badPlatforms = [

      # CMake Error at cmake/cpu_extension.cmake:78 (find_isa):

      # find_isa Function invoked with incorrect arguments for function named: