Add model like runner workflow (huggingface/transformers)
The Add model like runner workflow from huggingface/transformers, explained and optimized by Latchkey.
- C
- CI health grade
- 69/100
- Health score
- 1
- Job per trigger
- 3
- Fixes in the optimized version
In short
- CI health: C (fair)
- The optimized version below points runs-on at Latchkey managed runners and adds run de-duplication and job timeouts.
What it does
This is the Add model like runner workflow from the huggingface/transformers repository, a real project running GitHub Actions. It is shown here with attribution under its Apache-2.0 license.
Below, Latchkey shows a faster, safer version produced by its optimization engine.
The workflow
name: Add model like runner
on:
push:
branches:
- none # put main here when this is fixed
#pull_request:
# paths:
# - "src/**"
# - "tests/**"
# - ".github/**"
# types: [opened, synchronize, reopened]
permissions:
contents: read
jobs:
run_tests_templates_like:
name: "Add new model like template tests"
runs-on: ubuntu-22.04
steps:
- uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
with:
persist-credentials: false
- name: Install dependencies
run: |
sudo apt -y update && sudo apt install -y libsndfile1-dev
- name: Load cached virtual environment
uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
id: cache
with:
path: ~/venv/
key: v4-tests_model_like-${{ hashFiles('setup.py') }}
- name: Create virtual environment on cache miss
if: steps.cache.outputs.cache-hit != 'true'
run: |
python -m venv ~/venv && . ~/venv/bin/activate
pip install --upgrade pip!=21.3
pip install -e .[dev]
- name: Check transformers location
# make `transformers` available as package (required since we use `-e` flag) and check it's indeed from the repo.
run: |
. ~/venv/bin/activate
python setup.py develop
transformers_install=$(pip list -e | grep transformers)
transformers_install_array=($transformers_install)
transformers_loc=${transformers_install_array[-1]}
transformers_repo_loc=$(pwd .)
if [ "$transformers_loc" != "$transformers_repo_loc" ]; then
echo "transformers is from $transformers_loc but it shoud be from $transformers_repo_loc/src."
echo "A fix is required. Stop testing."
exit 1
fi
- name: Create model files
run: |
. ~/venv/bin/activate
transformers add-new-model-like --config_file tests/fixtures/add_distilbert_like_config.json --path_to_repo .
make style
make fix-copies
- name: Run all PyTorch modeling test
run: |
. ~/venv/bin/activate
python -m pytest -n 2 --dist=loadfile -s --make-reports=tests_new_models tests/bert_new/test_modeling_bert_new.py
- name: Run style changes
run: |
. ~/venv/bin/activate
make style && make quality && make repo-consistency
- name: Failure short reports
if: ${{ always() }}
run: cat reports/tests_new_models/failures_short.txt
- name: Test suite reports artifacts
if: ${{ always() }}
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
with:
name: run_all_tests_new_models_test_reports
path: reports/tests_new_modelsThe same workflow, on Latchkey
Removes redundant runs and caps runaway jobs. Added and changed lines are highlighted.
name: Add model like runner
on:
push:
branches:
- none # put main here when this is fixed
#pull_request:
# paths:
# - "src/**"
# - "tests/**"
# - ".github/**"
# types: [opened, synchronize, reopened]
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
jobs:
run_tests_templates_like:
timeout-minutes: 30
name: "Add new model like template tests"
runs-on: ubuntu-22.04
runs-on: latchkey-small
steps:
- uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
with:
persist-credentials: false
- name: Install dependencies
run: |
sudo apt -y update && sudo apt install -y libsndfile1-dev
- name: Load cached virtual environment
uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
id: cache
with:
path: ~/venv/
key: v4-tests_model_like-${{ hashFiles('setup.py') }}
- name: Create virtual environment on cache miss
if: steps.cache.outputs.cache-hit != 'true'
run: |
python -m venv ~/venv && . ~/venv/bin/activate
pip install --upgrade pip!=21.3
pip install -e .[dev]
- name: Check transformers location
# make `transformers` available as package (required since we use `-e` flag) and check it's indeed from the repo.
run: |
. ~/venv/bin/activate
python setup.py develop
transformers_install=$(pip list -e | grep transformers)
transformers_install_array=($transformers_install)
transformers_loc=${transformers_install_array[-1]}
transformers_repo_loc=$(pwd .)
if [ "$transformers_loc" != "$transformers_repo_loc" ]; then
echo "transformers is from $transformers_loc but it shoud be from $transformers_repo_loc/src."
echo "A fix is required. Stop testing."
exit 1
fi
- name: Create model files
run: |
. ~/venv/bin/activate
transformers add-new-model-like --config_file tests/fixtures/add_distilbert_like_config.json --path_to_repo .
make style
make fix-copies
- name: Run all PyTorch modeling test
run: |
. ~/venv/bin/activate
python -m pytest -n 2 --dist=loadfile -s --make-reports=tests_new_models tests/bert_new/test_modeling_bert_new.py
- name: Run style changes
run: |
. ~/venv/bin/activate
make style && make quality && make repo-consistency
- name: Failure short reports
if: ${{ always() }}
run: cat reports/tests_new_models/failures_short.txt
- name: Test suite reports artifacts
if: ${{ always() }}
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
with:
name: run_all_tests_new_models_test_reports
path: reports/tests_new_models
What changed
- Run on Latchkey managed runners with one line (
runs-on), which apply the fixes below automatically and self-heal transient failures. Each Linux x64 job moves to the Latchkey size matching its vCPU count,latchkey-smallfor a standard 2 vCPU runner; macOS, Windows and arm64 jobs stay as written. - Cancel superseded runs when a branch or PR gets a newer push.
- Add a job timeout so a hung step cannot burn hours of runner time.
What Latchkey heals here
This workflow has steps that commonly fail on transient issues (network, registries, flaky browsers). On Latchkey managed runners they are detected, retried, and self-healed instead of failing your build:
- Dependency installs
This workflow runs 1 job per trigger. It runs on a standard GitHub-hosted Linux x64 runner, which in a private repository bills $0.006 a minute against $0.0025 on latchkey-small, 58% lower; in a public repository like this one it is free. Estimate the monthly difference with your own run counts.
Actions used in this workflow
- actions/checkout
- actions/cache
- actions/upload-artifact