DeepSpeedPrecisionPlugin

class pytorch_lightning.plugins.precision.DeepSpeedPrecisionPlugin(precision, amp_type, amp_level=None)[source]

Bases: pytorch_lightning.plugins.precision.precision_plugin.PrecisionPlugin

Precision plugin for DeepSpeed integration.

backward(model, closure_loss, *args, **kwargs)[source]

Performs the actual backpropagation.

Parameters

• model (LightningModule) – the model to be optimized

• closure_loss (Tensor) – the loss value obtained from the closure

• optimizer – current optimizer being used. None if using manual optimization

Return type

None

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

DeepSpeed handles gradient clipping internally.

Return type

None

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

Hook to run the optimizer step.

Return type

Any