Nightly Regression Tests #63
This file contains hidden or 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
| # Copyright 2026 Tensor Auto Inc. All rights reserved. | |
| # | |
| # 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. | |
| name: Nightly Regression Tests | |
| on: | |
| schedule: | |
| # Run at 2:00 AM PST every day (10:00 AM UTC) | |
| - cron: '0 10 * * *' | |
| workflow_dispatch: | |
| permissions: | |
| contents: read | |
| env: | |
| PYTORCH_CUDA_ALLOC_CONF: "expandable_segments:True" | |
| MUJOCO_GL: "egl" | |
| PYOPENGL_PLATFORM: "egl" | |
| jobs: | |
| start-runner: | |
| name: Start GPU Runner | |
| runs-on: ubuntu-latest | |
| permissions: | |
| id-token: write | |
| contents: read | |
| steps: | |
| - name: Configure AWS Credentials | |
| uses: aws-actions/configure-aws-credentials@v4 | |
| with: | |
| role-to-assume: ${{ secrets.AWS_ROLE_ARN }} | |
| aws-region: us-west-2 | |
| - name: Start Instance | |
| run: | | |
| aws autoscaling set-desired-capacity --auto-scaling-group-name github-runner-asg-g6-12xlarge --desired-capacity 1 | |
| echo "Waiting for instance to be ready..." | |
| train-regression: | |
| name: Train with Model Parallelism | |
| needs: start-runner | |
| runs-on: [g6.12xlarge] | |
| timeout-minutes: 30 | |
| container: | |
| image: nvidia/cuda:12.2.0-devel-ubuntu22.04 | |
| options: --gpus all --ipc=host | |
| steps: | |
| - name: Checkout code | |
| uses: actions/checkout@v4 | |
| with: | |
| persist-credentials: false | |
| - name: Install system dependencies | |
| run: | | |
| apt-get update && apt-get install -y python3 python3-pip git ffmpeg libegl1 libegl-mesa0 libegl-dev libgl1 libglx-mesa0 libgles2 mesa-utils curl cmake build-essential | |
| - name: Install uv | |
| uses: astral-sh/setup-uv@v5 | |
| with: | |
| version: "latest" | |
| enable-cache: false | |
| - name: Install dependencies | |
| run: | | |
| uv sync --extra dev --extra libero | |
| - name: Check GPU | |
| run: nvidia-smi | |
| - name: Set up HuggingFace authentication | |
| shell: bash | |
| env: | |
| HF_TOKEN: ${{ secrets.HF_TOKEN }} | |
| run: | | |
| source .venv/bin/activate | |
| huggingface-cli login --token $HF_TOKEN | |
| - name: Set up Wandb authentication | |
| shell: bash | |
| env: | |
| WANDB_API_KEY: ${{ secrets.WANDB_API_KEY }} | |
| run: | | |
| source .venv/bin/activate | |
| wandb login $WANDB_API_KEY | |
| wandb offline | |
| - name: Set up Libero Configs | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| mkdir -p /tmp/libero-assets/libero/libero | |
| export LIBERO_CONFIG_PATH="$(pwd)/.github/assets/libero" | |
| - name: Run Training | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| opentau-train --accelerate-config=configs/examples/accelerate_deepspeed_config.yaml --config_path=configs/dev/ci_config.json --output_dir=outputs/train/ci/ 2>&1 | tee /tmp/train.log | |
| - name: Check Loss Drop | |
| id: check-loss-drop | |
| continue-on-error: true | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| python3 .github/scripts/check_loss_drop.py --log_path=/tmp/train.log --expected_length=25 | |
| echo "Loss drop confirmed" | |
| - name: Check Non-Zero Grad Norm | |
| id: check-grad-norm | |
| continue-on-error: true | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| python3 .github/scripts/check_nonzero_grad_norm.py --log_path=/tmp/train.log --expected_length=25 | |
| echo "Non-zero grad norm confirmed" | |
| - name: Check Accumulate Grad Sync | |
| id: check-grad-sync | |
| continue-on-error: true | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| python3 .github/scripts/check_accumulate_grad_sync.py --log_path=/tmp/train.log --expected_length=50 | |
| echo "Accumulate grad sync confirmed" | |
| - name: Check State Keys | |
| id: check-state-keys | |
| continue-on-error: true | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| python3 .github/scripts/check_state_keys.py --log_path=/tmp/train.log --source=hf | |
| echo "Checks for state keys passed" | |
| - name: Convert Checkpoint | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| ./src/opentau/scripts/convert_checkpoint.sh outputs/train/ci/checkpoints/000025 | |
| # - name: Resume Training | |
| # shell: bash | |
| # run: | | |
| # source .venv/bin/activate | |
| # opentau-train --accelerate-config=configs/examples/accelerate_deepspeed_config.yaml --config_path=outputs/train/ci/checkpoints/000025/train_config.json --resume=true --steps=50 2>&1 | tee /tmp/resume.log | |
| # - name: Check Loss Drop (after Resume) | |
| # continue-on-error: true | |
| # shell: bash | |
| # run: | | |
| # source .venv/bin/activate | |
| # python3 .github/scripts/check_loss_drop.py --log_path=/tmp/train.log --expected_length=25 --resume_log_path=/tmp/resume.log --resume_expected_length=25 | |
| # echo "Loss drop confirmed" | |
| # - name: Check Non-Zero Grad Norm (after Resume) | |
| # continue-on-error: true | |
| # shell: bash | |
| # run: | | |
| # source .venv/bin/activate | |
| # python3 .github/scripts/check_nonzero_grad_norm.py --log_path=/tmp/resume.log --expected_length=25 | |
| # echo "Non-zero grad norm confirmed" | |
| # - name: Check Accumulate Grad Sync (after Resume) | |
| # continue-on-error: true | |
| # shell: bash | |
| # run: | | |
| # source .venv/bin/activate | |
| # python3 .github/scripts/check_accumulate_grad_sync.py --log_path=/tmp/resume.log --expected_length=50 | |
| # echo "Accumulate grad sync confirmed" | |
| # - name: Check State Keys (after Resume) | |
| # continue-on-error: true | |
| # shell: bash | |
| # run: | | |
| # source .venv/bin/activate | |
| # python3 .github/scripts/check_state_keys.py --log_path=/tmp/resume.log --source=local | |
| # echo "Checks for state keys passed" | |
| - name: Run Inference | |
| shell: bash | |
| run: | | |
| source .venv/bin/activate | |
| python src/opentau/scripts/inference.py --config_path=outputs/train/ci/checkpoints/000025/train_config.json | |
| - name: Fail if checks failed | |
| if: always() | |
| env: | |
| LOSS_DROP: ${{ steps.check-loss-drop.outcome }} | |
| GRAD_NORM: ${{ steps.check-grad-norm.outcome }} | |
| GRAD_SYNC: ${{ steps.check-grad-sync.outcome }} | |
| STATE_KEYS: ${{ steps.check-state-keys.outcome }} | |
| run: | | |
| failed="" | |
| [ "$LOSS_DROP" == "failure" ] && failed="$failed check-loss-drop" | |
| [ "$GRAD_NORM" == "failure" ] && failed="$failed check-grad-norm" | |
| [ "$GRAD_SYNC" == "failure" ] && failed="$failed check-grad-sync" | |
| [ "$STATE_KEYS" == "failure" ] && failed="$failed check-state-keys" | |
| if [ -n "$failed" ]; then | |
| echo "The following checks failed:$failed" | |
| exit 1 | |
| fi | |
| stop-runner: | |
| name: Stop GPU Runner | |
| needs: [start-runner, train-regression] | |
| if: always() | |
| runs-on: ubuntu-latest | |
| permissions: | |
| id-token: write | |
| contents: read | |
| steps: | |
| - name: Configure AWS Credentials | |
| uses: aws-actions/configure-aws-credentials@v4 | |
| with: | |
| role-to-assume: ${{ secrets.AWS_ROLE_ARN }} | |
| aws-region: us-west-2 | |
| - name: Stop Instance | |
| run: | | |
| aws autoscaling set-desired-capacity --auto-scaling-group-name github-runner-asg-g6-12xlarge --desired-capacity 0 |