neural_compressor.torch.utils.environ

Intel Neural Compressor PyTorch environment check.

Functions

is_ipex_imported(→ bool)	Check whether intel_extension_for_pytorch is imported.
is_transformers_imported(→ bool)	Check whether transformers is imported.
is_package_available(package_name)	Check if the package exists in the environment without importing.
is_hpex_available()	Returns whether hpex is available.
is_optimum_available()	Return whether optimum-habana is available.
is_optimum_habana_available()	Return whether optimum-habana is available.
is_ipex_available()	Return whether ipex is available.
get_ipex_version()	Return ipex version if ipex exists.
get_torch_version()	Return torch version if ipex exists.
get_accelerator([device_name])	Return the recommended accelerator based on device priority.
device_synchronize(raw_func)	Function decorator that calls accelerated.synchronize before and after a function call.

Module Contents

neural_compressor.torch.utils.environ.is_ipex_imported() → bool

Check whether intel_extension_for_pytorch is imported.

neural_compressor.torch.utils.environ.is_transformers_imported() → bool

Check whether transformers is imported.

neural_compressor.torch.utils.environ.is_package_available(package_name)

Check if the package exists in the environment without importing.

Parameters:

package_name (str) – package name

neural_compressor.torch.utils.environ.is_hpex_available()

Returns whether hpex is available.

neural_compressor.torch.utils.environ.is_optimum_available()

Return whether optimum-habana is available.

neural_compressor.torch.utils.environ.is_optimum_habana_available()

Return whether optimum-habana is available.

neural_compressor.torch.utils.environ.is_ipex_available()

Return whether ipex is available.

neural_compressor.torch.utils.environ.get_ipex_version()

Return ipex version if ipex exists.

neural_compressor.torch.utils.environ.get_torch_version()

Return torch version if ipex exists.

neural_compressor.torch.utils.environ.get_accelerator(device_name='auto')

Return the recommended accelerator based on device priority.

neural_compressor.torch.utils.environ.device_synchronize(raw_func)

Function decorator that calls accelerated.synchronize before and after a function call.