Skip to content

Add get_logits method and NLLB tokenizer #1360

Add get_logits method and NLLB tokenizer

Add get_logits method and NLLB tokenizer #1360

Workflow file for this run

name: Continuous integration
on:
push:
branches:
- main
paths-ignore:
- '**.md'
- 'CITATION.cff'
- 'LICENSE'
- '.gitignore'
- 'docs/**'
pull_request:
branches:
- main
paths-ignore:
- '**.md'
- 'CITATION.cff'
- 'LICENSE'
- '.gitignore'
- 'docs/**'
workflow_dispatch:
inputs:
manual_revision_reference:
required: false
type: string
manual_revision_test:
required: false
type: string
env:
REVISION_REFERENCE: v2.8.2
#9d31b2ec4df6d8228f370ff20c8267ec6ba39383 earliest compatible v2.7.0 + pretrained_hf param
jobs:
Tests:
strategy:
matrix:
os: [ ubuntu-latest ] #, macos-latest ]
python: [ 3.8 ]
job_num: [ 4 ]
job: [ 1, 2, 3, 4 ]
runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v3
with:
fetch-depth: 0
ref: ${{ inputs.manual_revision_test }}
- name: Set up Python ${{ matrix.python }}
id: pythonsetup
uses: actions/setup-python@v4
with:
python-version: ${{ matrix.python }}
- name: Venv cache
id: venv-cache
uses: actions/cache@v3
with:
path: .env
key: venv-${{ matrix.os }}-${{ steps.pythonsetup.outputs.python-version }}-${{ hashFiles('requirements*') }}
- name: Pytest durations cache
uses: actions/cache@v3
with:
path: .test_durations
key: test_durations-${{ matrix.os }}-${{ steps.pythonsetup.outputs.python-version }}-${{ matrix.job }}-${{ github.run_id }}
restore-keys: test_durations-0-
- name: Setup
if: steps.venv-cache.outputs.cache-hit != 'true'
run: |
python3 -m venv .env
source .env/bin/activate
make install
make install-test
make install-training
- name: Prepare test data
run: |
source .env/bin/activate
python -m pytest \
--quiet --co \
--splitting-algorithm least_duration \
--splits ${{ matrix.job_num }} \
--group ${{ matrix.job }} \
-m regression_test \
tests \
| head -n -2 | grep -Po 'test_inference_with_data\[\K[^]]*(?=-False]|-True])' \
> models_gh_runner.txt
if [ -n "${{ inputs.manual_revision_reference }}" ]; then
REVISION_REFERENCE=${{ inputs.manual_revision_reference }}
fi
python tests/util_test.py \
--save_model_list models_gh_runner.txt \
--model_list models_gh_runner.txt \
--git_revision $REVISION_REFERENCE
- name: Unit tests
run: |
source .env/bin/activate
if [[ -f .test_durations ]]
then
cp .test_durations durations_1
mv .test_durations durations_2
fi
python -m pytest \
-x -s -v \
--splitting-algorithm least_duration \
--splits ${{ matrix.job_num }} \
--group ${{ matrix.job }} \
--store-durations \
--durations-path durations_1 \
--clean-durations \
-m "not regression_test" \
tests
OPEN_CLIP_TEST_REG_MODELS=models_gh_runner.txt python -m pytest \
-x -s -v \
--store-durations \
--durations-path durations_2 \
--clean-durations \
-m "regression_test" \
tests
jq -s -S 'add' durations_* > .test_durations
- name: Collect pytest durations
uses: actions/upload-artifact@v3
with:
name: pytest_durations_${{ matrix.os }}-${{ matrix.python }}-${{ matrix.job }}
path: .test_durations
Collect:
needs: Tests
runs-on: ubuntu-latest
steps:
- name: Cache
uses: actions/cache@v3
with:
path: .test_durations
key: test_durations-0-${{ github.run_id }}
- name: Collect
uses: actions/download-artifact@v3
with:
path: artifacts
- name: Consolidate
run: |
jq -n -S \
'reduce (inputs | to_entries[]) as {$key, $value} ({}; .[$key] += $value)' \
artifacts/pytest_durations_*/.test_durations > .test_durations