neural_compressor.torch.utils.auto_accelerator

Auto Accelerator Module.

Classes

`INCAcceleratorType`	Create a collection of name/value pairs.
`AcceleratorRegistry`	Accelerator Registry.
`Auto_Accelerator`	Auto Accelerator Base class.
`CPU_Accelerator`	CPU Accelerator.
`CUDA_Accelerator`	CUDA Accelerator.
`XPU_Accelerator`	XPU Accelerator.
`HPU_Accelerator`	HPU Accelerator.

Functions

`register_accelerator`(→ Callable[Ellipsis, Any])	Register new accelerator.
`auto_detect_accelerator`(→ Auto_Accelerator)	Automatically detects and selects the appropriate accelerator.

Module Contents

class neural_compressor.torch.utils.auto_accelerator.INCAcceleratorType(*args, **kwds)[source]

Create a collection of name/value pairs.

Example enumeration:

>>> class Color(Enum):
...     RED = 1
...     BLUE = 2
...     GREEN = 3

Access them by:

attribute access:
```
>>> Color.RED
<Color.RED: 1>
```
value lookup:
```
>>> Color(1)
<Color.RED: 1>
```
name lookup:
```
>>> Color['RED']
<Color.RED: 1>
```

Enumerations can be iterated over, and know how many members they have:

>>> len(Color)
3

>>> list(Color)
[<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]

Methods can be added to enumerations, and members can have their own attributes – see the documentation for details.

class neural_compressor.torch.utils.auto_accelerator.AcceleratorRegistry[source]: Accelerator Registry.

neural_compressor.torch.utils.auto_accelerator.register_accelerator(name: str, priority: float = 0) → Callable[Ellipsis, Any][source]

Register new accelerator.

Usage example:: @register_accelerator(name=”cuda”, priority=100) class CUDA_Accelerator:

…

Parameters:

name – the accelerator name.
priority – the priority of the accelerator. A larger number indicates a higher priority,

class neural_compressor.torch.utils.auto_accelerator.Auto_Accelerator[source]: Auto Accelerator Base class.

class neural_compressor.torch.utils.auto_accelerator.CPU_Accelerator[source]: CPU Accelerator.

class neural_compressor.torch.utils.auto_accelerator.CUDA_Accelerator[source]: CUDA Accelerator.

class neural_compressor.torch.utils.auto_accelerator.XPU_Accelerator[source]: XPU Accelerator.

class neural_compressor.torch.utils.auto_accelerator.HPU_Accelerator[source]: HPU Accelerator.

neural_compressor.torch.utils.auto_accelerator.auto_detect_accelerator(device_name='auto') → Auto_Accelerator[source]

Automatically detects and selects the appropriate accelerator.

Force use the cpu on node has both cpu and gpu: INC_TARGET_DEVICE=cpu python main.py … The INC_TARGET_DEVICE is case insensitive. The environment variable INC_TARGET_DEVICE has higher priority than the device_name. TODO: refine the docs and logic later