Skip to content

Auto Parallel 8 GPU Integration Tests #128

Auto Parallel 8 GPU Integration Tests

Auto Parallel 8 GPU Integration Tests #128

name: Auto Parallel 8 GPU Integration Tests
on:
push:
branches: [ main ]
tags:
- ciflow/8gpu/*
paths:
- 'torchtitan/experiments/autoparallel/**'
- '.github/workflows/integration_test_8gpu_autoparallel.yaml'
pull_request:
paths:
- 'torchtitan/experiments/autoparallel/**'
- '.github/workflows/integration_test_8gpu_autoparallel.yaml'
schedule:
# Runs every 12 hours
- cron: '0 */12 * * *'
concurrency:
group: unit-test${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && github.run_number || github.ref }}
cancel-in-progress: true
defaults:
run:
shell: bash -l -eo pipefail {0}
permissions:
id-token: write
contents: read
jobs:
# Step 1: Dynamically compute the matrix based on conditions
set-matrix:
uses: ./.github/workflows/set-matrix.yaml
# Step 2: Use the dynamic matrix in the build-test job
build-test:
needs: set-matrix
uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
strategy:
fail-fast: false
matrix: ${{ fromJSON(needs.set-matrix.outputs.matrix) }}
with:
runner: ${{ matrix.runner }}
gpu-arch-type: ${{ matrix.gpu-arch-type }}
gpu-arch-version: ${{ matrix.gpu-arch-version }}
docker-image: ${{ matrix.docker-image }}
repository: pytorch/torchtitan
upload-artifact: outputs
timeout: 45
script: |
set -eux
# The generic Linux job chooses to use base env, not the one setup by the image
CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
conda activate "${CONDA_ENV}"
# Log GPU info / driver version for debugging.
if [[ "${{ matrix.gpu-arch-type }}" == "cuda" ]]; then
DRIVER_VERSION=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader | head -n 1 || true)
echo "CUDA driver version: ${DRIVER_VERSION}"
elif [[ "${{ matrix.gpu-arch-type }}" == "rocm" ]]; then
echo "ROCm info:"
rocminfo || true
fi
pip config --user set global.progress_bar off
python -m pip install --force-reinstall --pre torch --index-url ${{ matrix.index-url }}
# Install autoparallel - required dependency for autoparallel experiment
python -m pip install git+https://github.com/meta-pytorch/autoparallel.git
sudo mkdir -p "$RUNNER_TEMP/artifacts-to-be-uploaded"
sudo chown -R $(id -u):$(id -g) "$RUNNER_TEMP/artifacts-to-be-uploaded"
python -m torchtitan.experiments.autoparallel.tests.integration_tests --gpu_arch_type ${{ matrix.gpu-arch-type }} $RUNNER_TEMP/artifacts-to-be-uploaded --ngpu 4