Skip to content

Introduce GPU testing to JAX Triton #189

Introduce GPU testing to JAX Triton

Introduce GPU testing to JAX Triton #189

Workflow file for this run

name: ci
on:
push:
branches:
- main
pull_request:
branches:
- main
permissions:
contents: read # to fetch code
concurrency:
group: ${{ github.workflow }}-${{ github.head_ref || github.ref }}
cancel-in-progress: true
jobs:
lint:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332 # ratchet:actions/checkout@v4
- name: Set up Python 3.10
uses: actions/setup-python@f677139bbe7f9c59b41e40162b753c062f5d49a3 # ratchet:actions/setup-python@v5
with:
python-version: '3.10'
- uses: pre-commit/action@2c7b3805fd2a0fd8c1884dcaebf91fc102a13ecd # ratchet:pre-commit/[email protected]
test:
runs-on: linux-x86-g2-48-l4-4gpu
container:
# TODO: change image based on what is needed for these tests
image: index.docker.io/tensorflow/build@sha256:7fb38f0319bda36393cad7f40670aa22352b44421bb906f5cf34d543acd8e1d2 # ratchet:tensorflow/build:latest-python3.11
steps:
- uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332 # ratchet:actions/checkout@v4
- name: Setup Compat Driver
run: |
# This container should already have the CUDA apt repos setup
apt-get update
apt-get install -y --no-install-recommends cuda-compat-12-6
- name: Setup Released JAX
run: |
pip install -U "jax[cuda12]"
pip install pytest
- name: Test JAX Triton
run: |
echo "Running JAX Triton GPU Tests"
nvidia-smi
pip install .
# Need newer ml-dtypes because we install newer numpy
pip install --upgrade ml-dtypes
pytest -v --tb=short tests/