Advanced Configuration
The default settings for Intel® Extension for PyTorch* are sufficient for most use cases. However, if users want to customize Intel® Extension for PyTorch*, advanced configuration is available at build time and runtime.
Build Time Configuration
The following build options are supported by Intel® Extension for PyTorch*. Users who install Intel® Extension for PyTorch* via source compilation could override the default configuration by explicitly setting a build option ON or OFF, and then build.
| Build Option | Default Value |
Description |
|---|---|---|
| USE_ONEMKL | ON | Use oneMKL BLAS |
| USE_CHANNELS_LAST_1D | ON | Use channels last 1d |
| USE_PERSIST_STREAM | ON | Use persistent oneDNN stream |
| USE_SCRATCHPAD_MODE | ON | Use oneDNN scratchpad mode |
| USE_PRIMITIVE_CACHE | ON | Cache oneDNN primitives by FRAMEWORK for specific operators |
| USE_QUEUE_BARRIER | ON | Use queue submit_barrier, otherwise use dummy kernel |
| USE_MULTI_CONTEXT | OFF | Create DPC++ runtime context per device |
| USE_PROFILER | ON | USE XPU Profiler in build. |
| USE_SYCL_ASSERT | OFF | Enables assert in sycl kernel |
| USE_ITT_ANNOTATION | OFF | Enables ITT annotation in sycl kernel |
| USE_SPLIT_FP64_LOOPS | ON | Split FP64 loops into separate kernel for element-wise kernels |
| USE_XETLA | OFF | Use XeTLA based customer kernels |
| BUILD_BY_PER_KERNEL | OFF | Build by DPC++ per_kernel option (exclusive with USE_AOT_DEVLIST) |
| BUILD_INTERNAL_DEBUG | OFF | Use internal debug code path |
| BUILD_SEPARATE_OPS | OFF | Build each operator in separate library |
| BUILD_SIMPLE_TRACE | ON | Build simple trace for each registered operator |
| BUILD_JIT_QUANTIZATION_SAVE | OFF | Support jit quantization model save and load |
| USE_AOT_DEVLIST | "" | Set device list for AOT build |
| BUILD_OPT_LEVEL | "" | Add build option -Ox, accept values: 0/1 |
| BUILD_STATIC_ONEMKL | OFF if "USE_ONEMKL" is ON, otherwise OFF | Static link with oneMKL |
For above build options which can be configured to ON or OFF, users can configure them to 1 or 0 also, while ON equals to 1 and OFF equals to 0.
Runtime Configuration
The following launch options are supported in Intel® Extension for PyTorch*. Users who execute AI models on XPU could override the default configuration by explicitly setting the option value at runtime using environment variables, and then launch the execution.
| Launch Option CPU, GPU |
Default Value |
Description |
|---|---|---|
| IPEX_FP32_MATH_MODE | FP32 | Set values for FP32 math mode (valid values: FP32, TF32, BF32). Refer to API Documentation for details. |
| Launch Option GPU ONLY |
Default Value |
Description |
|---|---|---|
| IPEX_VERBOSE | 0 | Set verbose level with synchronization execution mode |
| IPEX_XPU_SYNC_MODE | 0 | Set 1 to enforce synchronization execution mode |
| IPEX_TILE_AS_DEVICE | 1 | Set 0 to disable tile partition and map per root device |
| Launch Option Experimental |
Default Value |
Description |
|---|---|---|
| IPEX_SIMPLE_TRACE | 0 | Set 1 to enable simple trace for all operators* |
For above launch options which can be configured to 1 or 0, users can configure them to ON or OFF also, while ON equals to 1 and OFF equals to 0.
Examples to configure the launch options:
Set one or more options before running the model
export IPEX_VERBOSE=1
export IPEX_FP32_MATH_MODE=TF32
...
python ResNet50.py
Set one option when running the model
IPEX_VERBOSE=1 python ResNet50.py
Set more than one options when running the model
IPEX_VERBOSE=1 IPEX_FP32_MATH_MODE=TF32 python ResNet50.py