scale fixes for resnet (#1506) #3825
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: Build and Test with AIE tools on Ryzen AI | |
on: | |
push: | |
branches: | |
- main | |
- test-ryzen-ai | |
pull_request: | |
merge_group: | |
# Allows you to run this workflow manually from the Actions tab by | |
# selecting CI and then "Run workflow" menu on the right branch | |
# and clicking on "launch_tmate_terminal_for_debug". | |
# Unfortunately this works only for the default branch. | |
# So you can either | |
# - change the default branch of the PR on the GitHub repository owning the PR | |
# and launching in Actions tab; | |
# - or edit directly the step below which runs tmate and push to the | |
# PR, ignoring the manual workflow launch. | |
workflow_dispatch: | |
launch_tmate_terminal_for_debug: | |
type: boolean | |
description: 'Run the build with tmate debugging enabled (https://github.com/marketplace/actions/debugging-with-tmate)' | |
required: false | |
default: false | |
defaults: | |
run: | |
shell: bash | |
concurrency: | |
# A PR number if a pull request and otherwise the commit hash. This cancels | |
# queued and in-progress runs for the same PR (presubmit) or commit | |
# (postsubmit). | |
group: ci-build-test-ryzenai-${{ github.event.number || github.sha }} | |
cancel-in-progress: true | |
env: | |
DEBIAN_FRONTEND: noninteractive | |
XILINXD_LICENSE_FILE: /opt/xilinx/Xilinx.lic | |
VITIS: /opt/Xilinx/Vitis/2023.2 | |
CMAKE_ARGS: | | |
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \ | |
-DCMAKE_C_COMPILER_LAUNCHER=ccache \ | |
-DCMAKE_EXE_LINKER_FLAGS_INIT="-fuse-ld=lld" \ | |
-DCMAKE_MODULE_LINKER_FLAGS_INIT="-fuse-ld=lld" \ | |
-DCMAKE_SHARED_LINKER_FLAGS_INIT="-fuse-ld=lld" \ | |
-DXRT_ROOT=/opt/xilinx/xrt \ | |
-DAIE_ENABLE_PYTHON_PASSES=OFF \ | |
-DAIE_ENABLE_XRT_PYTHON_BINDINGS=ON \ | |
-DAIE_INCLUDE_INTEGRATION_TESTS=OFF | |
# -j12 here to reduce the number of parallel chess jobs. | |
LIT_OPTS: -sv --time-tests -j12 --timeout 600 --show-unsupported --show-excluded | |
jobs: | |
build-tests: | |
name: Run Tests on Ryzen AI | |
runs-on: amd8700g | |
steps: | |
- uses: actions/checkout@v3 | |
with: | |
submodules: "true" | |
# Launch an ssh session via a proxy server if there is a need | |
# for debug. This seems to live for 35 min max | |
# https://github.com/mxschmitt/action-tmate | |
- name: Setup tmate session | |
uses: mxschmitt/action-tmate@v3 | |
# To run this, launch it manually on the default branch and | |
# click on "launch_tmate_terminal_for_debug" | |
if: github.event_name == 'workflow_dispatch' | |
&& inputs.launch_tmate_terminal_for_debug | |
- name: Run commands | |
run: | | |
pip cache purge | |
source /opt/xilinx/xrt/setup.sh | |
python -m venv aie-venv | |
source aie-venv/bin/activate | |
pip install -r python/requirements.txt | |
pip install -r python/requirements_ml.txt | |
pip install jupyter | |
sed -i.bak 's/OUTPUT_TIMEOUT = 10/OUTPUT_TIMEOUT = 100/g' \ | |
$(python -c 'import site; print(site.getsitepackages()[0])')/jupyter_client/runapp.py | |
VERSION=$(utils/clone-llvm.sh --get-wheel-version) | |
pip -q download mlir==$VERSION \ | |
-f https://github.com/Xilinx/mlir-aie/releases/expanded_assets/mlir-distro | |
unzip -q mlir-*.whl | |
# I have no clue why but the system clock on GHA containers is like 12 hours ahead. | |
# That means wheels have file with time stamps in the future which makes ninja loop | |
# forever when configuring. Set the time to some arbitrary stamp in the past just to be safe. | |
find mlir -exec touch -a -m -t 201108231405.14 {} \; | |
mkdir build | |
pushd build | |
export PATH=/opt/Xilinx/Vitis/2023.2/bin:/opt/Xilinx/Vitis/2023.2/aietools/bin:$PATH | |
cmake .. -G Ninja \ | |
-DPython3_EXECUTABLE=$(which python) \ | |
-DLLVM_EXTERNAL_LIT=$(which lit) \ | |
-DCMAKE_INSTALL_PREFIX=$PWD/../mlir_aie \ | |
-DCMAKE_MODULE_PATH=$PWD/../cmake/modulesXilinx \ | |
-DMLIR_DIR=$PWD/../mlir/lib/cmake/mlir \ | |
$CMAKE_ARGS | |
ninja install | |
ninja check-aie | |
popd | |
build-examples: | |
name: Run Examples on Ryzen AI | |
runs-on: amd7940hs | |
steps: | |
- uses: actions/checkout@v3 | |
with: | |
submodules: "true" | |
# Launch an ssh session via a proxy server if there is a need | |
# for debug. This seems to live for 35 min max | |
# https://github.com/mxschmitt/action-tmate | |
- name: Setup tmate session | |
uses: mxschmitt/action-tmate@v3 | |
# To run this, launch it manually on the default branch and | |
# click on "launch_tmate_terminal_for_debug" | |
if: github.event_name == 'workflow_dispatch' | |
&& inputs.launch_tmate_terminal_for_debug | |
- name: Run commands | |
run: | | |
pip cache purge | |
source /opt/xilinx/xrt/setup.sh | |
python -m venv aie-venv | |
source aie-venv/bin/activate | |
pip install -r python/requirements.txt | |
pip install -r python/requirements_ml.txt | |
pip install jupyter | |
sed -i.bak 's/OUTPUT_TIMEOUT = 10/OUTPUT_TIMEOUT = 100/g' \ | |
$(python -c 'import site; print(site.getsitepackages()[0])')/jupyter_client/runapp.py | |
VERSION=$(utils/clone-llvm.sh --get-wheel-version) | |
pip -q download mlir==$VERSION \ | |
-f https://github.com/Xilinx/mlir-aie/releases/expanded_assets/mlir-distro | |
unzip -q mlir-*.whl | |
# I have no clue why but the system clock on GHA containers is like 12 hours ahead. | |
# That means wheels have file with time stamps in the future which makes ninja loop | |
# forever when configuring. Set the time to some arbitrary stamp in the past just to be safe. | |
find mlir -exec touch -a -m -t 201108231405.14 {} \; | |
mkdir build | |
pushd build | |
export PATH=$VITIS/bin:$VITIS/aietools/bin:$PATH | |
cmake .. -G Ninja \ | |
-DPython3_EXECUTABLE=$(which python) \ | |
-DLLVM_EXTERNAL_LIT=$(which lit) \ | |
-DCMAKE_INSTALL_PREFIX=$PWD/../mlir_aie \ | |
-DCMAKE_MODULE_PATH=$PWD/../cmake/modulesXilinx \ | |
-DMLIR_DIR=$PWD/../mlir/lib/cmake/mlir \ | |
$CMAKE_ARGS | |
ninja install | |
ninja check-reference-designs | |
popd |