Skip to content

Cpu fused kernel

Cpu fused kernel #11

Workflow file for this run

name: PR Tests
on:
pull_request:
types: [opened, synchronize, reopened]
branches: [main]
paths:
- ".github/workflows/test-runner.yml"
- ".github/workflows/tests-pr.yml"
- ".github/scripts/build-cpu.sh"
- ".github/scripts/build-cuda.sh"
- "bitsandbytes/**"
- "csrc/**"
- "include/**"
- "tests/**"
- "CMakeLists.txt"
- "setup.py"
- "pyproject.toml"
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
cancel-in-progress: true
jobs:
test-cpu:
name: CPU
if: github.repository == 'bitsandbytes-foundation/bitsandbytes'
strategy:
fail-fast: false
matrix:
platform: [linux-x64, linux-aarch64, macos]
# default runners don't have AVX-512 support, but icelake does
cpu_type: ["", icelake]
torch_version: ["2.3.1", "2.9.1"]
exclude:
# aarch64 minimum torch version is 2.5.1
- platform: linux-aarch64
torch_version: "2.3.1"
# icelake only applies to linux-x64
- platform: linux-aarch64
cpu_type: icelake
- platform: macos
cpu_type: icelake
include:
# Add aarch64 with torch 2.5.1 instead of 2.3.1
- platform: linux-aarch64
cpu_type: ""
torch_version: "2.5.1"
uses: ./.github/workflows/test-runner.yml
with:
platform: ${{ matrix.platform }}
backend: cpu
torch_version: ${{ matrix.torch_version }}
pypi_index: "https://download.pytorch.org/whl/cpu"
cpu_type: ${{ matrix.cpu_type }}
test-cuda:
name: CUDA
if: github.repository == 'bitsandbytes-foundation/bitsandbytes'
strategy:
fail-fast: false
matrix:
platform: [linux-x64]
gpu_type: [T4, L40S]
cuda_version: ["11.8.0", "12.8.1", "13.0.1"]
include:
# Map CUDA version to torch version and PyPI index
- cuda_version: "11.8.0"
torch_version: "2.3.1"
pypi_index: "https://download.pytorch.org/whl/cu118"
- cuda_version: "12.8.1"
torch_version: "2.8.0"
pypi_index: "https://download.pytorch.org/whl/cu128"
- cuda_version: "13.0.1"
torch_version: "2.9.1"
pypi_index: "https://download.pytorch.org/whl/cu130"
# Windows CUDA test - single configuration
- platform: windows
gpu_type: T4
cuda_version: "11.8.0"
torch_version: "2.7.1"
pypi_index: "https://download.pytorch.org/whl/cu118"
uses: ./.github/workflows/test-runner.yml
with:
platform: ${{ matrix.platform }}
backend: cuda
cuda_version: ${{ matrix.cuda_version }}
gpu_type: ${{ matrix.gpu_type }}
torch_version: ${{ matrix.torch_version }}
pypi_index: ${{ matrix.pypi_index }}