Skip to content
.github/workflows/ci_gpu.yml

Fix DistributedDP Optimizer for Fast Gradient Clipping (#662) #27

Sign in to view logs

Fix DistributedDP Optimizer for Fast Gradient Clipping (#662)

Fix DistributedDP Optimizer for Fast Gradient Clipping (#662) #27

Workflow file for this run

.github/workflows/ci_gpu.yml at 6a674f9
name: CI_GPU

Check failure on line 1 in .github/workflows/ci_gpu.yml

View workflow run for this annotation

GitHub Actions / .github/workflows/ci_gpu.yml

Invalid workflow file 

`unittest_multi_gpu` is not a valid event name
on:
push:
branches:
- main
pull_request:
branches:
- main
unittest_multi_gpu:
runs-on: linux.4xlarge.nvidia.gpu
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Set up Python
uses: actions/setup-python@v2
with:
python-version: 3.9
- name: Install dependencies
run: |
./scripts/install_via_pip.sh -c
- name: Run multi-GPU unit tests
run: |
nvidia-smi
nvcc --version
python -m unittest opacus.tests.multigpu_gradcheck.GradientComputationTest.test_gradient_correct
integrationtest_py39_torch_release_cuda:
runs-on: ubuntu-latest
container:
# https://hub.docker.com/r/nvidia/cuda
image: nvidia/cuda:12.3.1-base-ubuntu22.04
options: --gpus all
env:
TZ: 'UTC'
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Set up Python
uses: actions/setup-python@v2
with:
python-version: 3.9
- name: Install dependencies
run: |
python -m pip install --upgrade pip
pip install pytest coverage coveralls
./scripts/install_via_pip.sh -c
- name: Install CUDA toolkit and cuDNN
run: |
apt-get update
apt-get install -y --no-install-recommends \
cuda-toolkit-11-1 \
libcudnn8=8.1.1.33-1+cuda11.1 \
libcudnn8-dev=8.1.1.33-1+cuda11.1
- name: Run MNIST integration test (CUDA)
run: |
mkdir -p runs/mnist/data
mkdir -p runs/mnist/test-reports
python examples/mnist.py --lr 0.25 --sigma 0.7 -c 1.5 --batch-size 64 --epochs 1 --data-root runs/mnist/data --n-runs 1 --device cuda
python -c "import torch; accuracy = torch.load('run_results_mnist_0.25_0.7_1.5_64_1.pt'); exit(0) if (accuracy[0]>0.78 and accuracy[0]<0.95) else exit(1)"
- name: Store MNIST test results
uses: actions/upload-artifact@v2
with:
name: mnist-gpu-reports
path: runs/mnist/test-reports
- name: Run CIFAR10 integration test (CUDA)
run: |
mkdir -p runs/cifar10/data
mkdir -p runs/cifar10/logs
mkdir -p runs/cifar10/test-reports
pip install tensorboard
python examples/cifar10.py --lr 0.1 --sigma 1.5 -c 10 --batch-size 2000 --epochs 10 --data-root runs/cifar10/data --log-dir runs/cifar10/logs --device cuda
python -c "import torch; model = torch.load('model_best.pth.tar'); exit(0) if (model['best_acc1']>0.4 and model['best_acc1']<0.49) else exit(1)"
python examples/cifar10.py --lr 0.1 --sigma 1.5 -c 10 --batch-size 2000 --epochs 10 --data-root runs/cifar10/data --log-dir runs/cifar10/logs --device cuda --grad_sample_mode no_op
python -c "import torch; model = torch.load('model_best.pth.tar'); exit(0) if (model['best_acc1']>0.4 and model['best_acc1']<0.49) else exit(1)"
- name: Store CIFAR10 test results
uses: actions/upload-artifact@v2
with:
name: cifar10-gpu-reports
path: runs/cifar10/test-reports
- name: Run IMDb integration test (CUDA)
run: |
mkdir -p runs/imdb/data
mkdir -p runs/imdb/test-reports
pip install --user datasets transformers
python examples/imdb.py --lr 0.02 --sigma 1.0 -c 1.0 --batch-size 64 --max-sequence-length 256 --epochs 2 --data-root runs/imdb/data --device cuda
python -c "import torch; accuracy = torch.load('run_results_imdb_classification.pt'); exit(0) if (accuracy>0.54 and accuracy<0.66) else exit(1)"
- name: Store IMDb test results
uses: actions/upload-artifact@v2
with:
name: imdb-gpu-reports
path: runs/imdb/test-reports
- name: Run charlstm integration test (CUDA)
run: |
mkdir -p runs/charlstm/data
wget https://download.pytorch.org/tutorial/data.zip -O runs/charlstm/data/data.zip
unzip runs/charlstm/data/data.zip -d runs/charlstm/data
rm runs/charlstm/data/data.zip
mkdir -p runs/charlstm/test-reports
pip install scikit-learn
python examples/char-lstm-classification.py --epochs=20 --learning-rate=2.0 --hidden-size=128 --delta=8e-5 --batch-size 400 --n-layers=1 --sigma=1.0 --max-per-sample-grad-norm=1.5 --data-root="runs/charlstm/data/data/names/" --device cuda --test-every 5
python -c "import torch; accuracy = torch.load('run_results_chr_lstm_classification.pt'); exit(0) if (accuracy>0.60 and accuracy<0.80) else exit(1)"
- name: Store test results
uses: actions/upload-artifact@v2
with:
name: charlstm-gpu-reports
path: runs/charlstm/test-reports
micro_benchmarks_py39_torch_release_cuda:
runs-on: ubuntu-latest
needs: [integrationtest_py39_torch_release_cuda]
container:
# https://hub.docker.com/r/nvidia/cuda
image: nvidia/cuda:12.3.1-base-ubuntu22.04
options: --gpus all
env:
TZ: 'UTC'
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Set up Python
uses: actions/setup-python@v2
with:
python-version: 3.9
- name: Install dependencies
run: |
python -m pip install --upgrade pip
pip install pytest coverage coveralls
./scripts/install_via_pip.sh
- name: Install CUDA toolkit and cuDNN
run: |
apt-get update
apt-get install -y --no-install-recommends \
cuda-toolkit-11-1 \
libcudnn8=8.1.1.33-1+cuda11.1 \
libcudnn8-dev=8.1.1.33-1+cuda11.1
- name: Run benchmark integration tests (CUDA)
run: |
mkdir -p benchmarks/results/raw
python benchmarks/run_benchmarks.py --batch_size 16 --layers "groupnorm instancenorm layernorm" --config_file ./benchmarks/config.json --root ./benchmarks/results/raw/ --cont
IFS=$' ';layers=("groupnorm" "instancenorm" "layernorm"); rm -rf /tmp/report_layers; mkdir -p /tmp/report_layers; IFS=$'\n'; files=`( echo "${layers[*]}" ) | sed 's/.*/.\/benchmarks\/results\/raw\/&*/'`
cp -v ${files[@]} /tmp/report_layers
report_id=`IFS=$'-'; echo "${layers[*]}"`
python benchmarks/generate_report.py --path-to-results /tmp/report_layers --save-path benchmarks/results/report-${report_id}.csv --format csv
python benchmarks/generate_report.py --path-to-results /tmp/report_layers --save-path benchmarks/results/report-${report_id}.pkl --format pkl
python benchmarks/check_threshold.py --report-path "./benchmarks/results/report-"$report_id".pkl" --metric runtime --threshold 3.0 --column "hooks/baseline"
python benchmarks/check_threshold.py --report-path "./benchmarks/results/report-"$report_id".pkl" --metric memory --threshold 1.6 --column "hooks/baseline"
- name: Store artifacts
uses: actions/upload-artifact@v2
with:
name: benchmarks-reports
path: benchmarks/results/