Interface OperationConfig.Builder
- All Superinterfaces:
Buildable,CopyableBuilder<OperationConfig.Builder,,OperationConfig> SdkBuilder<OperationConfig.Builder,,OperationConfig> SdkPojo
- Enclosing class:
OperationConfig
-
Method Summary
Modifier and TypeMethodDescriptiondefault OperationConfig.Builderforward(Consumer<ForwardConfig.Builder> forward) The configuration for a forward-only pass through the model without gradient computation.forward(ForwardConfig forward) The configuration for a forward-only pass through the model without gradient computation.default OperationConfig.BuilderforwardBackward(Consumer<ForwardBackwardConfig.Builder> forwardBackward) The configuration for a forward and backward pass that computes gradients.forwardBackward(ForwardBackwardConfig forwardBackward) The configuration for a forward and backward pass that computes gradients.default OperationConfig.BuilderloadState(Consumer<LoadStateConfig.Builder> loadState) The configuration for loading training state from a previously saved checkpoint.loadState(LoadStateConfig loadState) The configuration for loading training state from a previously saved checkpoint.default OperationConfig.BuilderoptimizerStep(Consumer<OptimizerStepConfig.Builder> optimizerStep) The configuration for an optimizer step that updates model weights using accumulated gradients.optimizerStep(OptimizerStepConfig optimizerStep) The configuration for an optimizer step that updates model weights using accumulated gradients.default OperationConfig.BuildersaveState(Consumer<SaveStateConfig.Builder> saveState) The configuration for saving full training state, including model weights and optimizer state.saveState(SaveStateConfig saveState) The configuration for saving full training state, including model weights and optimizer state.default OperationConfig.BuildersaveWeights(Consumer<SaveWeightsConfig.Builder> saveWeights) The configuration for saving model weights as a checkpoint.saveWeights(SaveWeightsConfig saveWeights) The configuration for saving model weights as a checkpoint.Methods inherited from interface software.amazon.awssdk.utils.builder.CopyableBuilder
copyMethods inherited from interface software.amazon.awssdk.utils.builder.SdkBuilder
applyMutation, buildMethods inherited from interface software.amazon.awssdk.core.SdkPojo
equalsBySdkFields, sdkFieldNameToField, sdkFields
-
Method Details
-
forward
The configuration for a forward-only pass through the model without gradient computation.
- Parameters:
forward- The configuration for a forward-only pass through the model without gradient computation.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
forward
The configuration for a forward-only pass through the model without gradient computation.
This is a convenience method that creates an instance of theForwardConfig.Builderavoiding the need to create one manually viaForwardConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed toforward(ForwardConfig).- Parameters:
forward- a consumer that will call methods onForwardConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-
forwardBackward
The configuration for a forward and backward pass that computes gradients.
- Parameters:
forwardBackward- The configuration for a forward and backward pass that computes gradients.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
forwardBackward
default OperationConfig.Builder forwardBackward(Consumer<ForwardBackwardConfig.Builder> forwardBackward) The configuration for a forward and backward pass that computes gradients.
This is a convenience method that creates an instance of theForwardBackwardConfig.Builderavoiding the need to create one manually viaForwardBackwardConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed toforwardBackward(ForwardBackwardConfig).- Parameters:
forwardBackward- a consumer that will call methods onForwardBackwardConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-
optimizerStep
The configuration for an optimizer step that updates model weights using accumulated gradients.
- Parameters:
optimizerStep- The configuration for an optimizer step that updates model weights using accumulated gradients.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
optimizerStep
The configuration for an optimizer step that updates model weights using accumulated gradients.
This is a convenience method that creates an instance of theOptimizerStepConfig.Builderavoiding the need to create one manually viaOptimizerStepConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed tooptimizerStep(OptimizerStepConfig).- Parameters:
optimizerStep- a consumer that will call methods onOptimizerStepConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-
saveWeights
The configuration for saving model weights as a checkpoint.
- Parameters:
saveWeights- The configuration for saving model weights as a checkpoint.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
saveWeights
The configuration for saving model weights as a checkpoint.
This is a convenience method that creates an instance of theSaveWeightsConfig.Builderavoiding the need to create one manually viaSaveWeightsConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed tosaveWeights(SaveWeightsConfig).- Parameters:
saveWeights- a consumer that will call methods onSaveWeightsConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-
saveState
The configuration for saving full training state, including model weights and optimizer state.
- Parameters:
saveState- The configuration for saving full training state, including model weights and optimizer state.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
saveState
The configuration for saving full training state, including model weights and optimizer state.
This is a convenience method that creates an instance of theSaveStateConfig.Builderavoiding the need to create one manually viaSaveStateConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed tosaveState(SaveStateConfig).- Parameters:
saveState- a consumer that will call methods onSaveStateConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-
loadState
The configuration for loading training state from a previously saved checkpoint.
- Parameters:
loadState- The configuration for loading training state from a previously saved checkpoint.- Returns:
- Returns a reference to this object so that method calls can be chained together.
-
loadState
The configuration for loading training state from a previously saved checkpoint.
This is a convenience method that creates an instance of theLoadStateConfig.Builderavoiding the need to create one manually viaLoadStateConfig.builder().When the
Consumercompletes,SdkBuilder.build()is called immediately and its result is passed toloadState(LoadStateConfig).- Parameters:
loadState- a consumer that will call methods onLoadStateConfig.Builder- Returns:
- Returns a reference to this object so that method calls can be chained together.
- See Also:
-