Accelerator

class pytorch_lightning.accelerators.Accelerator[source]

Bases: abc.ABC

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

Currently there are accelerators for:

• CPU

• GPU

• TPU

• IPU

• HPU

abstract static auto_device_count()[source]

Get the device count when set to auto.

Return type:

int

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.

Return type:

Any

abstract static is_available()[source]

Detect if the hardware is available.

Return type:

bool

abstract static parse_devices(devices)[source]

Accelerator device parsing logic.

Return type:

Any

setup(trainer)[source]

Setup plugins for the trainer fit and creates optimizers.

Parameters:

trainer (Trainer) – the trainer instance

Return type:

None

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.

Return type:

None

teardown()[source]

Clean up any state created by the accelerator.

Return type:

None