MixedPrecision — PyTorch Lightning 2.6.0 documentation (original) (raw)

class lightning.pytorch.plugins.precision.MixedPrecision(precision, device, scaler=None)[source]¶

Bases: Precision

Plugin for Automatic Mixed Precision (AMP) training with torch.autocast.

Parameters:

• precision¶ (Literal['16-mixed', 'bf16-mixed']) – Whether to use torch.float16 ('16-mixed') or torch.bfloat16 ('bf16-mixed').
• device¶ (str) – The device for torch.autocast.
• scaler¶ (Optional[GradScaler]) – An optional torch.cuda.amp.GradScaler to use.

clip_gradients(optimizer, clip_val=0.0, gradient_clip_algorithm=GradClipAlgorithmType.NORM)[source]¶

Clips the gradients.

Return type:

None

forward_context()[source]¶

Enable autocast context.

Return type:

Generator[None, None, None]

load_state_dict(state_dict)[source]¶

Called when loading a checkpoint, implement to reload precision plugin state given precision plugin state_dict.

Parameters:

state_dict¶ (dict[str, Any]) – the precision plugin state returned by state_dict.

Return type:

None

optimizer_step(optimizer, model, closure, **kwargs)[source]¶

Hook to run the optimizer step.

Return type:

Any

pre_backward(tensor, module)[source]¶

Runs before precision plugin executes backward.

Parameters:

• tensor¶ (Tensor) – The tensor that will be used for backpropagation
• module¶ (LightningModule) – The module that was involved in producing the tensor and whose parameters need the gradients

Return type:

Tensor

state_dict()[source]¶

Called when saving a checkpoint, implement to generate precision plugin state_dict.

Return type:

dict[str, Any]

Returns:

A dictionary containing precision plugin state.