Skip to content

GPU CI Fix (Pin runs-on GPU image) #84

GPU CI Fix (Pin runs-on GPU image)

GPU CI Fix (Pin runs-on GPU image) #84

Workflow file for this run

name: "tests"
on: [push, pull_request, workflow_dispatch]
concurrency:
group: build-${{ github.head_ref || github.run_id }}
cancel-in-progress: true
jobs:
cpu-ci:
name: CPU unit tests and build
runs-on: ubuntu-20.04
steps:
- name: Checkout Git Repository
uses: actions/checkout@v3
with:
submodules: recursive
- name: Free additional space on runner
run: ./.github/workflows/helpers/free_space_on_runner.sh
- name: Install nix
uses: cachix/install-nix-action@v25
with:
github_access_token: '${{ secrets.GITHUB_TOKEN }}'
- uses: cachix/cachix-action@v14
with:
name: ff
skipPush: true
# authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
- name: setup nix develop shell
uses: nicknovitski/[email protected]
env:
NIXPKGS_ALLOW_UNFREE: 1
with:
arguments: ".#ci --accept-flake-config"
- name: ccache
uses: hendrikmuhs/[email protected]
- name: Regenerate all dtgen files
run: |
proj dtgen --force
- name: Run cmake
run: |
proj cmake --dtgen-skip
- name: Run build and tests
run: |
proj test --dtgen-skip -j$(nproc) --coverage --skip-gpu-tests
- name: Upload code coverage
uses: codecov/codecov-action@v4
with:
token: ${{ secrets.CODECOV_TOKEN }}
file: main_coverage.info
flags: unittests
name: codecov-umbrella
fail_ci_if_error: false
verbose: true
gpu-ci:
name: GPU unit tests
# needs: cpu-ci
runs-on:
- runs-on
- family=g4dn.xlarge
- image=runs-on-gpu-pinned
strategy:
max-parallel: 1
fail-fast: false
steps:
- name: checkout git repository
uses: actions/checkout@v3
with:
submodules: recursive
# - name: remove the module system
# run: |
# rm /etc/profile.d/modules.sh
- name: mount ephemeral drive to nix
run: |
sudo mkdir $HOME/nix && sudo mkdir /nix && sudo mount --bind $HOME/nix /nix
# Note that this step is not only necessary to make space for the nix-provided
# dependencies, but also to prevent contention between the nix-provided nccl and the
# system-provided ones in the DL AMI
- name: free additional space on runner
run: ./.github/workflows/helpers/free_space_on_runner_gpu.sh
- name: run nvidia-smi
run: |
nvidia-smi
- name: show disk usage
run: |
du --all --human-readable --threshold 1G .[^.]* /* || true
- name: install nix
uses: cachix/install-nix-action@v25
with:
github_access_token: '${{ secrets.GITHUB_TOKEN }}'
- uses: cachix/cachix-action@v14
with:
name: ff
skipPush: true
# authToken: '${{ secrets.CACHIX_AUTH_TOKEN }}'
- name: setup nix develop shell
uses: lockshaw/nix-develop@116f24a2996eb68cfadddb5657b4b923e5f69ec5
env:
NIXPKGS_ALLOW_UNFREE: 1
with:
arguments: ".#gpu-ci --accept-flake-config --impure"
- name: ccache
uses: hendrikmuhs/[email protected]
- name: regenerate all dtgen files
run: |
proj dtgen --force
- name: run cmake
run: |
proj cmake --dtgen-skip
- name: build and run gpu tests
run: |
proj test --dtgen-skip -j$(nproc) --skip-build-cpu-tests