Accelerator¶

class pytorch_lightning.accelerators.Accelerator[source]¶

The Accelerator Base Class. An Accelerator is meant to deal with one type of Hardware.

Currently there are accelerators for:

abstract static auto_device_count()[source]¶

Get the device count when set to auto.

get_device_stats(device)[source]¶

Get stats for a given device.

Parameters: device¶ (Union[str, device]) – device for which to get stats
Return type: Dict[str, Any]
Returns: Dictionary of device stats

abstract static get_parallel_devices(devices)[source]¶

Gets parallel devices for the Accelerator.

abstract static is_available()[source]¶

Detect if the hardware is available.

abstract static parse_devices(devices)[source]¶

Accelerator device parsing logic.

setup(trainer)[source]¶

Setup plugins for the trainer fit and creates optimizers.

setup_environment(root_device)[source]¶

Setup any processes or distributed connections.

This is called before the LightningModule/DataModule setup hook which allows the user to access the accelerator environment before setup is complete.

Clean up any state created by the accelerator.