CIFAR-10 benchmark CML #29
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: CIFAR-10 benchmark CML | |
on: | |
workflow_dispatch: | |
inputs: | |
git-ref: | |
description: Repo reference (branch, tag or SHA) | |
default: "main" | |
required: true | |
type: string | |
benchmark: | |
description: Benchmark to run (cifar-10-8b or cifar-10-16b) | |
default: "cifar-10-16b" | |
type: choice | |
options: | |
- "cifar-10-8b" | |
- "cifar-10-16b" | |
instance_type: | |
description: Instance type on which to launch benchmarks | |
default: "m6i.metal" | |
type: choice | |
options: | |
- "m6i.metal" | |
- "u-6tb1.112xlarge" | |
num_samples: | |
description: Number of samples to use | |
default: "3" | |
type: string | |
required: true | |
# FIXME: Add recurrent launching | |
# https://github.com/zama-ai/concrete-ml-internal/issues/1851 | |
# Global environnement variables | |
env: | |
# Github action url (used by slack notification) | |
ACTION_RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} | |
AGENT_TOOLSDIRECTORY: /opt/hostedtoolcache | |
RUNNER_TOOL_CACHE: /opt/hostedtoolcache | |
# Jobs | |
jobs: | |
start-cifar-runner: | |
name: Launch AWS instances | |
runs-on: ubuntu-20.04 | |
defaults: | |
run: | |
shell: bash | |
container: | |
image: ubuntu:20.04 | |
outputs: | |
label: ${{ steps.start-cifar10-8bit-runner.outputs.label }} | |
ec2-instance-id: ${{ steps.start-cifar10-8bit-runner.outputs.ec2-instance-id || '' }} | |
steps: | |
- name: Configure AWS credentials | |
uses: aws-actions/configure-aws-credentials@010d0da01d0b5a38af31e9c3470dbfdabdecca3a | |
with: | |
aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }} | |
aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }} | |
aws-region: ${{ secrets.AWS_REGION }} | |
- name: Start CIFAR-10 8-bit runner | |
id: start-cifar10-8bit-runner | |
uses: machulav/ec2-github-runner@2c4d1dcf2c54673ed3bfd194c4b6919ed396a209 | |
with: | |
mode: start | |
github-token: ${{ secrets.EC2_RUNNER_BOT_TOKEN }} | |
ec2-image-id: ${{ secrets.AWS_EC2_AMI }} | |
ec2-instance-type: ${{ github.event.inputs.instance_type }} | |
subnet-id: ${{ secrets.AWS_EC2_SUBNET_ID }} | |
security-group-id: ${{ secrets.AWS_EC2_SECURITY_GROUP_ID }} | |
aws-resource-tags: > | |
[ | |
{"Key": "Name", "Value": "cml-benchmark-cifar10"}, | |
{"Key": "GitHubRepository", "Value": "${{ github.repository }}"}, | |
{"Key": "Actor", "Value": "${{ github.actor }}"}, | |
{"Key": "Action", "Value": "${{ github.action }}"}, | |
{"Key": "GitHash", "Value": "${{ github.sha }}"}, | |
{"Key": "RefName", "Value": "${{ github.ref_name }}"}, | |
{"Key": "RunId", "Value": "${{ github.run_id }}"}, | |
{"Key": "Team", "Value": "CML"} | |
] | |
run-cifar-10: | |
needs: [start-cifar-runner] | |
name: Run benchmark | |
runs-on: ${{ needs.start-cifar-runner.outputs.label }} | |
env: | |
PIP_INDEX_URL: ${{ secrets.PIP_INDEX_URL }} | |
PIP_EXTRA_INDEX_URL: ${{ secrets.PIP_EXTRA_INDEX_URL }} | |
steps: | |
- name: Add masks | |
run: | | |
echo "::add-mask::${{ secrets.INTERNAL_PYPI_URL_FOR_MASK }}" | |
echo "::add-mask::${{ secrets.INTERNAL_REPO_URL_FOR_MASK }}" | |
echo "::add-mask::${{ secrets.INTERNAL_PYPI_URL }}" | |
echo "::add-mask::${{ secrets.INTERNAL_REPO_URL }}" | |
- name: Checkout code | |
uses: actions/checkout@b4ffde65f46336ab88eb53be808477a3936bae11 | |
with: | |
lfs: true | |
ref: ${{ github.event.inputs.git-ref }} | |
- name: Set up Python | |
uses: actions/setup-python@0a5c61591373683505ea898e09a3ea4f39ef2b9c | |
with: | |
python-version: "3.8" | |
- name: Install dependencies | |
id: install-deps | |
run: | | |
apt update | |
apt install --no-install-recommends -y gnome-keyring | |
apt install -y graphviz* graphviz-dev libgraphviz-dev pkg-config python3-dev | |
apt-mark hold docker.io | |
apt upgrade curl | |
./script/make_utils/setup_os_deps.sh | |
make setup_env | |
# CIFAR-10-8b benchmark | |
- name: Benchmark - CIFAR-10-8b | |
if: github.event.inputs.benchmark == 'cifar-10-8b' | |
run: | | |
source .venv/bin/activate | |
NUM_SAMPLES=${{ github.event.inputs.num_samples }} python3 ./use_case_examples/cifar/cifar_brevitas_with_model_splitting/infer_fhe.py | |
python3 ./benchmarks/convert_cifar.py --model-name "8-bit-split-v0" | |
# CIFAR-10-16b benchmark | |
- name: Benchmark - CIFAR-10-16b | |
if: github.event.inputs.benchmark == 'cifar-10-16b' | |
run: | | |
source .venv/bin/activate | |
NUM_SAMPLES=${{ github.event.inputs.num_samples }} python3 ./use_case_examples/cifar/cifar_brevitas_training/evaluate_one_example_fhe.py | |
python3 ./benchmarks/convert_cifar.py --model-name "16-bits-trained-v0" | |
- name: Archive raw predictions | |
uses: actions/upload-artifact@v4 | |
with: | |
name: predictions.csv | |
path: inference_results.csv | |
- name: Archive metrics | |
uses: actions/upload-artifact@v4 | |
with: | |
name: metrics.json | |
path: to_upload.json | |
- name: Archive MLIR | |
uses: actions/upload-artifact@v4 | |
with: | |
name: mlir.txt | |
path: cifar10.mlir | |
- name: Archive Graph | |
uses: actions/upload-artifact@v4 | |
with: | |
name: graph.txt | |
path: cifar10.graph | |
- name: Archive client | |
uses: actions/upload-artifact@v4 | |
with: | |
name: client.zip | |
path: client_server/client.zip | |
- name: Archive server | |
uses: actions/upload-artifact@v4 | |
with: | |
name: server.zip | |
path: client_server/server.zip | |
# We need to keep this as the last step to avoid not uploading the artifacts | |
# if the step crashes | |
- name: Upload results | |
id: upload-results | |
run: | | |
# Log the json | |
cat to_upload.json | jq | |
# We need to sleep to avoid log issues | |
sleep 1. | |
curl --version | |
# Upload the json to the benchmark database | |
curl --fail-with-body \ | |
-H "Authorization: Bearer ${{ secrets.NEW_ML_PROGRESS_TRACKER_TOKEN }}" \ | |
-H "Content-Type: application/json; charset=UTF-8" \ | |
--json @to_upload.json \ | |
-X POST "${{ secrets.NEW_ML_PROGRESS_TRACKER_URL }}experiment" | |
stop-runner: | |
name: Stop EC2 runner | |
needs: [run-cifar-10, start-cifar-runner] | |
runs-on: ubuntu-20.04 | |
timeout-minutes: 2 | |
if: ${{ always() }} | |
steps: | |
- name: Configure AWS credentials | |
uses: aws-actions/configure-aws-credentials@010d0da01d0b5a38af31e9c3470dbfdabdecca3a | |
with: | |
aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }} | |
aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }} | |
aws-region: ${{ secrets.AWS_REGION }} | |
- name: Stop EC2 runner | |
uses: machulav/ec2-github-runner@2c4d1dcf2c54673ed3bfd194c4b6919ed396a209 | |
if: ${{ always() }} | |
with: | |
mode: stop | |
github-token: ${{ secrets.EC2_RUNNER_BOT_TOKEN }} | |
label: ${{ needs.start-cifar-runner.outputs.label }} | |
ec2-instance-id: ${{ needs.start-cifar-runner.outputs.ec2-instance-id }} | |
slack-notification: | |
runs-on: ubuntu-20.04 | |
needs: [run-cifar-10] | |
steps: | |
- name: Slack Notification | |
if: ${{ always() }} | |
continue-on-error: true | |
uses: rtCamp/action-slack-notify@b24d75fe0e728a4bf9fc42ee217caa686d141ee8 | |
env: | |
SLACK_CHANNEL: ${{ secrets.SLACK_CHANNEL }} | |
SLACK_ICON: https://pbs.twimg.com/profile_images/1274014582265298945/OjBKP9kn_400x400.png | |
SLACK_COLOR: ${{ needs.run-cifar-10.result }} | |
SLACK_MESSAGE: "Benchmark action: ${{ github.event.inputs.benchmark }} (${{ env.ACTION_RUN_URL }}) ended with result: ${{ needs.run-cifar-10.result }}" | |
SLACK_USERNAME: ${{ secrets.BOT_USERNAME }} | |
SLACK_WEBHOOK: ${{ secrets.SLACK_WEBHOOK }} | |