UNet++ Inference best known configurations with Intel® Extension for PyTorch.
Use Case | Framework | Model Repo | Branch/Commit/Tag | Optional Patch |
---|---|---|---|---|
Inference | PyTorch | https://github.com/qubvel/segmentation_models.pytorch | - | - |
-
Host has Intel® Data Center GPU Flex Series
-
Host has installed latest Intel® Data Center GPU Flex Series Drivers https://dgpu-docs.intel.com/driver/installation.html
-
The following Intel® oneAPI Base Toolkit components are required:
- Intel® oneAPI DPC++ Compiler (Placeholder DPCPPROOT as its installation path)
- Intel® oneAPI Math Kernel Library (oneMKL) (Placeholder MKLROOT as its installation path)
- Intel® oneAPI MPI Library
- Intel® oneAPI TBB Library
Follow instructions at Intel® oneAPI Base Toolkit Download page to setup the package manager repository.
git clone https://github.com/IntelAI/models.git
cd models/models_v2/pytorch/unetpp/inference/gpu
- Create virtual environment
venv
and activate it:python3 -m venv venv . ./venv/bin/activate
- Run setup.sh
./setup.sh
- Install the latest GPU versions of torch, torchvision and intel_extension_for_pytorch:
python -m pip install torch==<torch_version> torchvision==<torchvision_version> intel-extension-for-pytorch==<ipex_version> --extra-index-url https://pytorch-extension.intel.com/release-whl-aitools/
- Set environment variables for Intel® oneAPI Base Toolkit:
Default installation location
{ONEAPI_ROOT}
is/opt/intel/oneapi
for root account,${HOME}/intel/oneapi
for other accountssource {ONEAPI_ROOT}/compiler/latest/env/vars.sh source {ONEAPI_ROOT}/mkl/latest/env/vars.sh source {ONEAPI_ROOT}/tbb/latest/env/vars.sh source {ONEAPI_ROOT}/mpi/latest/env/vars.sh source {ONEAPI_ROOT}/ccl/latest/env/vars.sh
- Setup required environment paramaters
Parameter | export command |
---|---|
MULTI_TILE | export MULTI_TILE=False (False) |
PLATFORM | export PLATFORM=Flex (Flex) |
OUTPUT_DIR | export OUTPUT_DIR=$PWD |
BATCH_SIZE (optional) | export BATCH_SIZE=8 |
PRECISION (optional) | export PRECISION=fp16 |
- Run
run_model.sh
Single-tile output will typically looks like:
Latency: 0.03823380470275879
Throughput: 209.23891990855813
Final results of the inference run can be found in results.yaml
file.
results:
- key: throughput
value: 209.23892
unit: fps
- key: latency
value: 0.03823380470275879
unit: s
- key: accuracy
value: None
unit: Acc