add Fuji v3 405b and solve HBM OOMs for larger models #12
Workflow file for this run
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: build-and-test | |
on: [pull_request, push] | |
jobs: | |
build-and-test-job: | |
runs-on: ubuntu-latest | |
strategy: | |
matrix: | |
test-group: [a, b, c, d, e] | |
steps: | |
- uses: actions/checkout@v4 | |
- uses: docker/setup-buildx-action@v3 | |
- name: Gather test files | |
run: find axlearn -name '*_test.py' > pytest_files.txt | |
- name: Split test files into groups | |
# GNU split lets us do "-n r/5" to round robin into 5 files without breaking lines | |
# BSD split requires knowing the number of lines and uses "-l XX" | |
run: split -n r/5 -a 1 pytest_files.txt split_pytest_files | |
- name: Select a test group | |
run: tr '\n' ' ' < split_pytest_files${{ matrix.test-group }} > test_files_oneline | |
- name: Read test inputs | |
id: test-selector | |
run: echo "PYTEST_FILES='$(cat test_files_oneline)'" >> "$GITHUB_OUTPUT" | |
- name: Run tests | |
uses: docker/build-push-action@v6 | |
with: | |
push: false | |
target: ci | |
context: . | |
build-args: | | |
SKIP_PRECOMMIT=--skip-pre-commit | |
PYTEST_FILES=${{ steps.test-selector.outputs.PYTEST_FILES }} |