CI: Combine conda and full testing into a single workflow (#14387)

* Remove conda job

* Remove conda job from readme

* Remove conda jobs from checkgroup

* Remove conda from docker builds

* Remove base-conda dockerfile

* Rewrite the strategy matrix while keeping equivalent

* Run the workflow on this branch

* Revert "Rewrite the strategy matrix while keeping equivalent"

This reverts commit e54298d60e57cffbf8107890987be3fe4a006c77.

* Add PyTorch versions

* Run on draft and disable unrelated costly CI

* Revert "Run the workflow on this branch"

This reverts commit 51ed8b905d8926b630dce4817124bd486135d3ec.

* tmp: Lightweight relevant CI

* Fix CI pathfilter

* Update matrix

* Drop skipping logic

* pip list

* reorder pip list

* tmp: lightweight ci

* Install specified pytorch

* Fix torch installation

* Uncomment steps

* Increase timeout

* bad merge

* Revert "Run on draft and disable unrelated costly CI"

This reverts commit eb5dc5e6bd.

* Update checkgroup

* Update docs and remove Python/PyTorch versions

* Remove pip-list

* Fail if wrong pytorch version installed

* Add Python 3.8, PyTorch 1.9 job

* tmp: remove azure jobs

* tmp: remove dockers

* tmp: remove others

* Run all combinations

* Include oldest

* Exclude no Python 3.10 distributions

* tmp: no concurrency

* tmp: double timeout

* Add pytest log reporter

* Add pytest-reportlog

* Fewer jobs

* Revert "tmp: no concurrency"

This reverts commit 4a7978dcb3.

* fix artifact name

* Revert test reports

* Revert unrelated changes

* Revert unrelated changes

* Add the combination of ex-conda jobs

* Update checkgroup

* revert timeout

* remove conda job

* revert docker build workflow file

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
This commit is contained in:
Akihiro Nitta 2022-09-30 11:39:04 +09:00 committed by GitHub
parent fdcb5cc90b
commit e47d5a2376
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 94 additions and 394 deletions

View File

@ -31,19 +31,23 @@ subprojects:
- "setup.cfg" # includes pytest config
checks:
# Note: updates here should be applied to the lightning_lite group
- "pl-conda (3.8, 1.10)"
- "pl-conda (3.8, 1.9)"
- "pl-conda (3.9, 1.11)"
- "pl-conda (3.9, 1.12)"
- "pl-cpu (macOS-11, 3.10, latest, stable)"
- "pl-cpu (macOS-11, 3.7, latest, stable)"
- "pl-cpu (macOS-11, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04, 3.10, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, oldest, stable)"
- "pl-cpu (windows-2022, 3.10, latest, stable)"
- "pl-cpu (windows-2022, 3.7, latest, stable)"
- "pl-cpu (windows-2022, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04 3.7, 1.9, oldest)"
- "pl-cpu (ubuntu-20.04 3.7, 1.12)"
- "pl-cpu (ubuntu-20.04 3.8, 1.9)"
- "pl-cpu (ubuntu-20.04 3.8, 1.10)"
- "pl-cpu (ubuntu-20.04 3.9, 1.11)"
- "pl-cpu (ubuntu-20.04 3.9, 1.12)"
- "pl-cpu (ubuntu-20.04 3.10, 1.12)"
- "pl-cpu (macos-11, 3.7, 1.9, oldest)"
- "pl-cpu (macos-11, 3.7, 1.12)"
- "pl-cpu (macos-11, 3.8, 1.10)"
- "pl-cpu (macos-11, 3.9, 1.11)"
- "pl-cpu (macos-11, 3.10, 1.12)"
- "pl-cpu (windows-2022, 3.7, 1.9, oldest)"
- "pl-cpu (windows-2022, 3.7, 1.12)"
- "pl-cpu (windows-2022, 3.8, 1.10)"
- "pl-cpu (windows-2022, 3.9, 1.11)"
- "pl-cpu (windows-2022, 3.10, 1.12)"
- "make-doctest (pytorch)"
- "make-html (pytorch)"
- "mypy"
@ -56,28 +60,27 @@ subprojects:
# TODO: since this job cannot run on forks, it cannot be required or it will block all PL PRs from forks
#- "test-on-tpus"
- id: "pytorch_lightning: Conda"
paths:
- ".github/workflows/ci-pytorch-test-conda.yml"
checks:
- "pl-conda (3.8, 1.10)"
- "pl-conda (3.8, 1.9)"
- "pl-conda (3.9, 1.11)"
- "pl-conda (3.9, 1.12)"
- id: "pytorch_lightning: CPU"
paths:
- ".github/workflows/ci-pytorch-test-full.yml"
checks:
- "pl-cpu (macOS-11, 3.10, latest, stable)"
- "pl-cpu (macOS-11, 3.7, latest, stable)"
- "pl-cpu (macOS-11, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04, 3.10, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, oldest, stable)"
- "pl-cpu (windows-2022, 3.10, latest, stable)"
- "pl-cpu (windows-2022, 3.7, latest, stable)"
- "pl-cpu (windows-2022, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04 3.7, 1.9, oldest)"
- "pl-cpu (ubuntu-20.04 3.7, 1.12)"
- "pl-cpu (ubuntu-20.04 3.8, 1.9)"
- "pl-cpu (ubuntu-20.04 3.8, 1.10)"
- "pl-cpu (ubuntu-20.04 3.9, 1.11)"
- "pl-cpu (ubuntu-20.04 3.9, 1.12)"
- "pl-cpu (ubuntu-20.04 3.10, 1.12)"
- "pl-cpu (macos-11, 3.7, 1.9, oldest)"
- "pl-cpu (macos-11, 3.7, 1.12)"
- "pl-cpu (macos-11, 3.8, 1.10)"
- "pl-cpu (macos-11, 3.9, 1.11)"
- "pl-cpu (macos-11, 3.10, 1.12)"
- "pl-cpu (windows-2022, 3.7, 1.9, oldest)"
- "pl-cpu (windows-2022, 3.7, 1.12)"
- "pl-cpu (windows-2022, 3.8, 1.10)"
- "pl-cpu (windows-2022, 3.9, 1.11)"
- "pl-cpu (windows-2022, 3.10, 1.12)"
- id: "pytorch_lightning: Slow"
paths:
@ -127,10 +130,6 @@ subprojects:
- ".github/workflows/*docker*.yml"
- "setup.py"
checks:
- "build-conda (3.8, 1.9, 11.1.1)"
- "build-conda (3.8, 1.10.1, 11.1.1)"
- "build-conda (3.9, 1.11, 11.3.1)"
- "build-conda (3.9, 1.12, 11.3.1)"
- "build-cuda (3.8, 1.9, 11.1.1)"
- "build-cuda (3.9, 1.10, 11.3.1)"
- "build-cuda (3.9, 1.11, 11.3.1)"
@ -166,15 +165,23 @@ subprojects:
- "lightning-lite (GPUs)"
- "mypy"
# Lite also requires PL checks as it depends on Lite
- "pl-cpu (macOS-11, 3.10, latest, stable)"
- "pl-cpu (macOS-11, 3.7, latest, stable)"
- "pl-cpu (macOS-11, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04, 3.10, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, latest, stable)"
- "pl-cpu (ubuntu-20.04, 3.7, oldest, stable)"
- "pl-cpu (windows-2022, 3.10, latest, stable)"
- "pl-cpu (windows-2022, 3.7, latest, stable)"
- "pl-cpu (windows-2022, 3.7, oldest, stable)"
- "pl-cpu (ubuntu-20.04 3.7, 1.9, oldest)"
- "pl-cpu (ubuntu-20.04 3.7, 1.12)"
- "pl-cpu (ubuntu-20.04 3.8, 1.9)"
- "pl-cpu (ubuntu-20.04 3.8, 1.10)"
- "pl-cpu (ubuntu-20.04 3.9, 1.11)"
- "pl-cpu (ubuntu-20.04 3.9, 1.12)"
- "pl-cpu (ubuntu-20.04 3.10, 1.12)"
- "pl-cpu (macos-11, 3.7, 1.9, oldest)"
- "pl-cpu (macos-11, 3.7, 1.12)"
- "pl-cpu (macos-11, 3.8, 1.10)"
- "pl-cpu (macos-11, 3.9, 1.11)"
- "pl-cpu (macos-11, 3.10, 1.12)"
- "pl-cpu (windows-2022, 3.7, 1.9, oldest)"
- "pl-cpu (windows-2022, 3.7, 1.12)"
- "pl-cpu (windows-2022, 3.8, 1.10)"
- "pl-cpu (windows-2022, 3.9, 1.11)"
- "pl-cpu (windows-2022, 3.10, 1.12)"
- "make-doctest (pytorch)"
- "make-html (pytorch)"
- "pytorch-lightning (GPUs)"

View File

@ -4,23 +4,25 @@
## Unit and Integration Testing
| workflow name | workflow file | action | accelerator\* | (Python, PyTorch) | OS |
| -------------------------- | ------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------- | ------------------------------------------------- | ------------------- |
| Test PyTorch full | .github/workflows/ci-pytorch-test-full.yml | Run all tests except for accelerator-specific, standalone and slow tests. | CPU | (3.7, 1.9), (3.7, 1.12), (3.9, 1.9), (3.9, 1.12) | linux, mac, windows |
| Test PyTorch with Conda | .github/workflows/ci-pytorch-test-conda.yml | Same as ci-pytorch-test-full.yml but with dependencies installed with conda. | CPU | (3.8, 1.9), (3.8, 1.10), (3.8, 1.11), (3.9, 1.12) | linux |
| Test slow | .github/workflows/ci-pytorch-test-slow.yml | Run only slow tests. Slow tests usually need to spawn threads and cannot be speed up or simplified. | CPU | (3.7, 1.11) | linux, mac, windows |
| pytorch-lightning (IPUs) | .azure-pipelines/ipu-tests.yml | Run only IPU-specific tests. | IPU | (3.8, 1.9) | linux |
| pytorch-lightning (HPUs) | .azure-pipelines/hpu-tests.yml | Run only HPU-specific tests. | HPU | (3.8, 1.10) | linux |
| pytorch-lightning (GPUs) | .azure-pipelines/gpu-tests.yml | Run all CPU and GPU-specific tests, standalone, and examples. Each standalone test needs to be run in separate processes to avoid unwanted interactions between test cases. | GPU | (3.9, 1.12) | linux |
| PyTorchLightning.Benchmark | .azure-pipelines/gpu-benchmark.yml | Run speed/memory benchmarks for parity with pure PyTorch. | GPU | (3.9, 1.12) | linux |
| test-on-tpus | .circleci/config.yml | Run only TPU-specific tests. | TPU | (3.7, 1.12) | linux |
| workflow name | workflow file | action | accelerator\* |
| -------------------------- | ------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------- |
| Test PyTorch full | .github/workflows/ci-pytorch-test-full.yml | Run all tests except for accelerator-specific, standalone and slow tests. | CPU |
| Test PyTorch slow | .github/workflows/ci-pytorch-test-slow.yml | Run only slow tests. Slow tests usually need to spawn threads and cannot be speed up or simplified. | CPU |
| pytorch-lightning (IPUs) | .azure-pipelines/ipu-tests.yml | Run only IPU-specific tests. | IPU |
| pytorch-lightning (HPUs) | .azure-pipelines/hpu-tests.yml | Run only HPU-specific tests. | HPU |
| pytorch-lightning (GPUs) | .azure-pipelines/gpu-tests.yml | Run all CPU and GPU-specific tests, standalone, and examples. Each standalone test needs to be run in separate processes to avoid unwanted interactions between test cases. | GPU |
| PyTorchLightning.Benchmark | .azure-pipelines/gpu-benchmark.yml | Run speed/memory benchmarks for parity with pure PyTorch. | GPU |
| test-on-tpus | .circleci/config.yml | Run only TPU-specific tests. | TPU |
- \*Accelerators used in CI
- GPU: 2 x NVIDIA Tesla V100
- TPU: Google GKE TPUv3
- IPU: [Colossus MK1 IPU](https://www.graphcore.ai/products/ipu)
- HPU: [Intel Habana Gaudi SYS-420GH-TNGR](https://www.supermicro.com/en/products/system/AI/4U/SYS-420GH-TNGR) which has 8 Gaudi accelerators
- To check which versions of Python or PyTorch are used for testing in our CI, see the corresponding workflow files or checkgroup cofig file at [`.github/checkgroup.yml`](../checkgroup.yml).
## Documentation
| workflow file | action |

View File

@ -127,44 +127,6 @@ jobs:
env:
SLACK_WEBHOOK_URL: ${{ secrets.SLACK_WEBHOOK_URL }}
build-conda:
runs-on: ubuntu-20.04
strategy:
fail-fast: false
matrix:
include:
- {python_version: "3.8", pytorch_version: "1.9", cuda_version: "11.1.1"}
- {python_version: "3.8", pytorch_version: "1.10.1", cuda_version: "11.1.1"}
- {python_version: "3.9", pytorch_version: "1.11", cuda_version: "11.3.1"}
- {python_version: "3.9", pytorch_version: "1.12", cuda_version: "11.3.1"}
steps:
- uses: actions/checkout@v3
- uses: docker/setup-buildx-action@v2
- uses: docker/login-action@v2
if: env.PUSH_TO_HUB == 'true'
with:
username: ${{ secrets.DOCKER_USERNAME }}
password: ${{ secrets.DOCKER_PASSWORD }}
- uses: docker/build-push-action@v3
with:
build-args: |
PYTHON_VERSION=${{ matrix.python_version }}
PYTORCH_VERSION=${{ matrix.pytorch_version }}
CUDA_VERSION=${{ matrix.cuda_version }}
file: dockers/base-conda/Dockerfile
push: ${{ env.PUSH_TO_HUB }}
tags: pytorchlightning/pytorch_lightning:base-conda-py${{ matrix.python_version }}-torch${{ matrix.pytorch_version }}
timeout-minutes: 95
- uses: ravsamhq/notify-slack-action@v2
if: failure() && env.PUSH_TO_HUB == 'true'
with:
status: ${{ job.status }}
token: ${{ secrets.GITHUB_TOKEN }}
notification_title: ${{ format('Conda; {0} py{1} for *{2}*', runner.os, matrix.python_version, matrix.pytorch_version) }}
message_format: '{emoji} *{workflow}* {status_message}, see <{run_url}|detail>, cc: <@U01A5T7EY9M>' # akihironitta
env:
SLACK_WEBHOOK_URL: ${{ secrets.SLACK_WEBHOOK_URL }}
build-ipu:
runs-on: ubuntu-20.04
strategy:

View File

@ -1,110 +0,0 @@
name: Test PyTorch with Conda
# see: https://help.github.com/en/actions/reference/events-that-trigger-workflows
on:
push:
branches: [master, "release/*"]
pull_request:
branches: [master, "release/*"]
paths:
- ".github/workflows/ci-pytorch-test-conda.yml"
- "requirements/pytorch/**"
- "src/pytorch_lightning/**"
- "tests/tests_pytorch/**"
- "setup.cfg" # includes pytest config
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref }}
cancel-in-progress: ${{ ! (github.ref == 'refs/heads/master' || startsWith(github.ref, 'refs/heads/release/')) }}
defaults:
run:
shell: bash -l {0}
jobs:
pl-conda:
runs-on: ubuntu-20.04
container: pytorchlightning/pytorch_lightning:base-conda-py${{ matrix.python-version }}-torch${{ matrix.pytorch-version }}
strategy:
fail-fast: false
matrix:
include:
- {python-version: "3.8", pytorch-version: "1.9"}
- {python-version: "3.8", pytorch-version: "1.10"}
- {python-version: "3.9", pytorch-version: "1.11"}
- {python-version: "3.9", pytorch-version: "1.12"}
timeout-minutes: 40
steps:
- name: Workaround for https://github.com/actions/checkout/issues/760
run: git config --global --add safe.directory /__w/lightning/lightning
- uses: actions/checkout@v3
- name: Update base dependencies
env:
PACKAGE_NAME: pytorch
FREEZE_REQUIREMENTS: 1
run: |
conda info
conda list
pip install -e .[test]
- name: Freeze PIL (hotfix)
# import of PILLOW_VERSION which they recently removed in v9.0 in favor of __version__
run: pip install "Pillow<9.0" # It messes with torchvision
- name: DocTests
working-directory: ./src
run: pytest pytorch_lightning --cov=pytorch_lightning
- name: Update all dependencies
env:
HOROVOD_BUILD_ARCH_FLAGS: "-mfma"
HOROVOD_WITHOUT_MXNET: 1
HOROVOD_WITHOUT_TENSORFLOW: 1
run: |
set -e
pip list
# adjust versions according installed Torch version
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/extra.txt
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/examples.txt
pip install -r requirements/pytorch/devel.txt -r requirements/pytorch/strategies.txt --find-links https://download.pytorch.org/whl/torch_stable.html
# set a per-test timeout of 2.5 minutes to fail sooner; this aids with hanging tests
pip install pytest-timeout
pip list
# sanity check
python requirements/pytorch/check-avail-extras.py
- name: Pull legacy checkpoints
run: bash .actions/pull_legacy_checkpoints.sh
- name: Testing PyTorch
working-directory: tests/tests_pytorch
run: coverage run --source pytorch_lightning -m pytest -v --timeout 150 --durations=50 --junitxml=results-${{ runner.os }}-torch${{ matrix.pytorch-version }}.xml
- name: Upload pytest results
uses: actions/upload-artifact@v3
with:
name: unittest-results-${{ runner.os }}-torch${{ matrix.pytorch-version }}
path: tests/tests_pytorch/results-${{ runner.os }}-torch${{ matrix.pytorch-version }}.xml
if: failure()
- name: Statistics
if: success()
working-directory: tests/tests_pytorch
run: |
coverage report
coverage xml
- name: Upload coverage to Codecov
uses: codecov/codecov-action@v3
if: success()
# see: https://github.com/actions/toolkit/issues/399
continue-on-error: true
with:
token: ${{ secrets.CODECOV_TOKEN }}
file: tests/tests_pytorch/coverage.xml
flags: cpu,pytest,torch${{ matrix.pytorch-version }}
name: CPU-coverage
fail_ci_if_error: false

View File

@ -29,17 +29,26 @@ jobs:
strategy:
fail-fast: false
matrix:
os: [ubuntu-20.04, windows-2022, macOS-11]
python-version: ["3.7", "3.10"] # minimum, maximum
requires: ["oldest", "latest"]
release: ["stable"]
exclude:
# There's no distribution of the oldest PyTorch 1.9 for Python 3.10.
# TODO: Remove the exclusion when dropping PyTorch 1.9 support.
- {python-version: "3.10", requires: "oldest"}
# TODO: re-enable RC testing
# include:
# - {os: ubuntu-20.04, python-version: "3.10", requires: "latest", release: "pre"}
include:
- {os: "ubuntu-20.04", python-version: "3.7", pytorch-version: "1.9", requires: "oldest"}
- {os: "ubuntu-20.04", python-version: "3.7", pytorch-version: "1.12"}
- {os: "ubuntu-20.04", python-version: "3.8", pytorch-version: "1.9"} # ex-conda
- {os: "ubuntu-20.04", python-version: "3.8", pytorch-version: "1.10"} # ex-conda
- {os: "ubuntu-20.04", python-version: "3.9", pytorch-version: "1.11"} # ex-conda
- {os: "ubuntu-20.04", python-version: "3.9", pytorch-version: "1.12"} # ex-conda
- {os: "ubuntu-20.04", python-version: "3.10", pytorch-version: "1.12"}
- {os: "macos-11", python-version: "3.7", pytorch-version: "1.9", requires: "oldest"}
- {os: "macos-11", python-version: "3.7", pytorch-version: "1.12"}
- {os: "macos-11", python-version: "3.8", pytorch-version: "1.10"}
- {os: "macos-11", python-version: "3.9", pytorch-version: "1.11"}
- {os: "macos-11", python-version: "3.10", pytorch-version: "1.12"}
- {os: "windows-2022", python-version: "3.7", pytorch-version: "1.9", requires: "oldest"}
- {os: "windows-2022", python-version: "3.7", pytorch-version: "1.12"}
- {os: "windows-2022", python-version: "3.8", pytorch-version: "1.10"}
- {os: "windows-2022", python-version: "3.9", pytorch-version: "1.11"}
- {os: "windows-2022", python-version: "3.10", pytorch-version: "1.12"}
# TODO: re-enable RC testing
# - {os: ubuntu-20.04, python-version: "3.10", release: "pre"}
timeout-minutes: 40
@ -59,7 +68,6 @@ jobs:
pip --version
pip install -q -r .actions/requirements.txt
# Github Actions: Run step on specific OS: https://stackoverflow.com/a/57948488/4521646
- name: Setup macOS
if: runner.os == 'macOS'
run: |
@ -75,8 +83,6 @@ jobs:
run: |
python .actions/assistant.py replace_oldest_ver
# Note: This uses an internal pip API and may not always work
# https://github.com/actions/cache/blob/master/examples.md#multiple-oss-in-a-workflow
- name: Get pip cache dir
id: pip-cache
run: echo "::set-output name=dir::$(pip cache dir)"
@ -97,9 +103,13 @@ jobs:
PACKAGE_NAME: pytorch
FREEZE_REQUIREMENTS: 1
run: |
flag=$(python -c "print('--pre' if '${{matrix.release}}' == 'pre' else '')" 2>&1)
url=$(python -c "print('test/cpu/torch_test.html' if '${{matrix.release}}' == 'pre' else 'cpu/torch_stable.html')" 2>&1)
pip install -e .[test] --upgrade $flag --find-links "https://download.pytorch.org/whl/${url}"
# adjust PyTorch versions in requirements files
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/base.txt ${{ matrix.pytorch-version }}
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/examples.txt ${{ matrix.pytorch-version }}
# install PL and optional dependencies for testing
pre_option=$(python -c "print('--pre' if '${{matrix.release}}' == 'pre' else '')" 2>&1)
url=https://download.pytorch.org/whl/$(python -c "print('test/cpu/torch_test.html' if '${{matrix.release}}' == 'pre' else 'cpu/torch_stable.html')" 2>&1)
pip install -e . -r requirements/pytorch/test.txt -r ./requirements/pytorch/extra.txt $pre_option -f ${url}
pip list
shell: bash
@ -107,14 +117,6 @@ jobs:
working-directory: ./src
run: pytest pytorch_lightning --cov=pytorch_lightning
- name: Install extra dependencies
run: |
# adjust versions according installed Torch version
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/extra.txt
pip install -r ./requirements/pytorch/extra.txt --find-links https://download.pytorch.org/whl/cpu/torch_stable.html --upgrade
pip list
shell: bash
- name: Reinstall Horovod if necessary
if: runner.os != 'windows'
env:
@ -130,6 +132,7 @@ jobs:
fi
horovodrun --check-build
python -c "import horovod.torch"
pip list
shell: bash
- name: Cache datasets
@ -139,7 +142,9 @@ jobs:
key: pl-dataset
- name: Sanity check
run: python requirements/pytorch/check-avail-extras.py
run: |
python -c "import torch; assert torch.__version__.startswith('${{ matrix.pytorch-version }}')"
python requirements/pytorch/check-avail-extras.py
- name: Testing Warnings
# the stacklevel can only be set on >=3.7
@ -160,12 +165,6 @@ jobs:
name: unittest-results-${{ runner.os }}-py${{ matrix.python-version }}-${{ matrix.requires }}-${{ matrix.release }}
path: tests/tests_pytorch/results-${{ runner.os }}-py${{ matrix.python-version }}-${{ matrix.requires }}-${{ matrix.release }}.xml
- name: Prepare Examples
run: |
# adjust versions according installed Torch version
python ./requirements/pytorch/adjust-versions.py requirements/pytorch/examples.txt
pip install -r requirements/pytorch/examples.txt --find-links https://download.pytorch.org/whl/cpu/torch_stable.html --upgrade
- name: Run Examples
working-directory: ./examples
run: python -m pytest test_pl_examples.py -v --durations=10

View File

@ -1,160 +0,0 @@
# Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
ARG CUDA_VERSION=11.3.1
FROM nvidia/cuda:${CUDA_VERSION}-devel-ubuntu20.04
ARG PYTHON_VERSION=3.9
ARG PYTORCH_VERSION=1.9
ARG CONDA_VERSION=4.11.0
SHELL ["/bin/bash", "-c"]
# https://techoverflow.net/2019/05/18/how-to-fix-configuring-tzdata-interactive-input-when-building-docker-images/
ENV \
PATH="$PATH:/root/.local/bin" \
DEBIAN_FRONTEND=noninteractive \
TZ=Europe/Prague \
# CUDA_TOOLKIT_ROOT_DIR="/usr/local/cuda" \
MKL_THREADING_LAYER=GNU
RUN \
# TODO: Remove the manual key installation once the base image is updated.
# https://github.com/NVIDIA/nvidia-docker/issues/1631
apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub && \
apt-get update -qq --fix-missing && \
NCCL_VER=$(dpkg -s libnccl2 | grep '^Version:' | awk -F ' ' '{print $2}' | awk -F '-' '{print $1}' | grep -ve '^\s*$') && \
CUDA_VERSION_MM="${CUDA_VERSION%.*}" && \
MAX_ALLOWED_NCCL=2.11.4 && \
TO_INSTALL_NCCL=$(echo -e "$MAX_ALLOWED_NCCL\n$NCCL_VER" | sort -V | head -n1)-1+cuda${CUDA_VERSION_MM} && \
apt-get install -y --no-install-recommends \
build-essential \
cmake \
git \
wget \
curl \
unzip \
ca-certificates \
libopenmpi-dev \
libnccl2=$TO_INSTALL_NCCL \
libnccl-dev=$TO_INSTALL_NCCL && \
# Install conda and python.
# NOTE new Conda does not forward the exit status... https://github.com/conda/conda/issues/8385
curl -o ~/miniconda.sh https://repo.anaconda.com/miniconda/Miniconda3-py38_${CONDA_VERSION}-Linux-x86_64.sh && \
chmod +x ~/miniconda.sh && \
~/miniconda.sh -b && \
rm ~/miniconda.sh && \
# Cleaning
apt-get autoremove -y && \
apt-get clean && \
rm -rf /root/.cache && \
rm -rf /var/lib/apt/lists/*
ENV \
PATH="/root/miniconda3/bin:$PATH" \
LD_LIBRARY_PATH="/root/miniconda3/lib:$LD_LIBRARY_PATH" \
CUDA_TOOLKIT_ROOT_DIR="/usr/local/cuda" \
MKL_THREADING_LAYER=GNU \
# MAKEFLAGS="-j$(nproc)" \
MAKEFLAGS="-j2" \
TORCH_CUDA_ARCH_LIST="3.7;5.0;6.0;7.0;7.5;8.0" \
CONDA_ENV=lightning
COPY environment.yml environment.yml
# conda init
RUN \
conda update -n base -c defaults conda && \
CUDA_VERSION_MM=$(python -c "print('.'.join('$CUDA_VERSION'.split('.')[:2]))") && \
conda create -y --name $CONDA_ENV \
python=${PYTHON_VERSION} pytorch=${PYTORCH_VERSION} torchvision cudatoolkit=${CUDA_VERSION_MM} \
-c nvidia -c pytorch -c pytorch-test && \
conda init bash && \
# NOTE: this requires that the channel is presented in the yaml before packages \
printf "import re;\nfname = 'environment.yml';\nreq = open(fname).read();\nfor n in ['python', 'pytorch', 'torchvision']:\n req = re.sub(rf'- {n}[>=]+', f'# - {n}=', req);\nopen(fname, 'w').write(req)" > prune.py && \
python prune.py && \
rm prune.py && \
cat environment.yml && \
conda env update --name $CONDA_ENV --file environment.yml && \
conda install "Pillow<9.0" && \
conda clean -ya && \
rm environment.yml
ENV \
PATH=/root/miniconda3/envs/${CONDA_ENV}/bin:$PATH \
LD_LIBRARY_PATH="/root/miniconda3/envs/${CONDA_ENV}/lib:$LD_LIBRARY_PATH"
COPY ./requirements/pytorch/ ./requirements/pytorch/
COPY ./.actions/assistant.py assistant.py
RUN \
pip list | grep torch && \
python -c "import torch; print(torch.__version__)" && \
pip install -q fire && \
python assistant.py requirements_prune_pkgs torch,torchvision && \
# Install remaining requirements
pip install --no-cache-dir -r requirements/pytorch/base.txt \
-r requirements/pytorch/extra.txt \
-r requirements/pytorch/examples.txt && \
rm assistant.py
ENV \
# if you want this environment to be the default o \ne, uncomment the following line:
CONDA_DEFAULT_ENV=${CONDA_ENV} \
HOROVOD_CUDA_HOME=$CUDA_TOOLKIT_ROOT_DIR \
HOROVOD_GPU_OPERATIONS=NCCL \
HOROVOD_WITH_PYTORCH=1 \
HOROVOD_WITHOUT_TENSORFLOW=1 \
HOROVOD_WITHOUT_MXNET=1 \
HOROVOD_WITH_GLOO=1 \
HOROVOD_WITH_MPI=1
RUN \
HOROVOD_BUILD_CUDA_CC_LIST=${TORCH_CUDA_ARCH_LIST//";"/","} && \
export HOROVOD_BUILD_CUDA_CC_LIST=${HOROVOD_BUILD_CUDA_CC_LIST//"."/""} && \
pip install --no-cache-dir -r requirements/pytorch/strategies.txt
RUN \
CUDA_VERSION_MAJOR=$(python -c "import torch ; print(torch.version.cuda.split('.')[0])") && \
py_ver=$(python -c "print(int('$PYTHON_VERSION'.split('.') >= '3.9'.split('.')))") && \
# install DALI, needed for examples
# todo: waiting for 1.4 - https://github.com/NVIDIA/DALI/issues/3144#issuecomment-877386691
if [ $py_ver -eq "0" ]; then \
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist "nvidia-dali-cuda${CUDA_VERSION_MAJOR}0>1.0" ; \
python -c 'from nvidia.dali.pipeline import Pipeline' ; \
fi
RUN \
# install NVIDIA apex
pip install --no-cache-dir --global-option="--cuda_ext" https://github.com/NVIDIA/apex/archive/refs/heads/master.zip && \
python -c "from apex import amp"
RUN \
# install Bagua
CUDA_VERSION_MM=$(python -c "print(''.join('$CUDA_VERSION'.split('.')[:2]))") && \
CUDA_VERSION_BAGUA=$(python -c "print([ver for ver in [116,113,111,102] if $CUDA_VERSION_MM >= ver][0])") && \
pip install "bagua-cuda$CUDA_VERSION_BAGUA" && \
if [[ "$CUDA_VERSION_MM" = "$CUDA_VERSION_BAGUA" ]]; then python -c "import bagua_core; bagua_core.install_deps()"; fi && \
python -c "import bagua; print(bagua.__version__)"
RUN \
# Show what we have
pip --version && \
conda info && \
pip list && \
python -c "import sys; ver = sys.version_info ; assert f'{ver.major}.{ver.minor}' == '$PYTHON_VERSION', ver" && \
python -c "import torch; assert torch.__version__.startswith('$PYTORCH_VERSION'), torch.__version__" && \
python requirements/pytorch/check-avail-extras.py && \
python requirements/pytorch/check-avail-strategies.py && \
rm -rf requirements/