Dspy.Module behaviour (dspy v0.1.0)

View Source

Behaviour for composable DSPy modules.

Modules are the building blocks of DSPy programs. They define forward passes for inference and can contain optimizable parameters.

Summary

Types

inputs()
outputs()
t()

Callbacks

forward(module, inputs)

Execute the module's forward pass.

parameters(module)

Get the module's optimizable parameters.

update_parameters(module, parameters)

Update the module with new parameters.

Functions

compose(modules)

Compose multiple modules in sequence.

forward(module, inputs)

Execute a module's forward pass.

parallel(modules)

Run modules in parallel and merge results.

parameters(module)

Get a module's parameters if it implements the callback.

update_parameters(module, parameters)

Update a module's parameters if it implements the callback.

Types

inputs()

@type inputs() :: map()

outputs()

@type outputs() :: Dspy.Prediction.t()

t()

@type t() :: struct()

Callbacks

forward(module, inputs)

@callback forward(module :: t(), inputs :: inputs()) :: {:ok, outputs()} | {:error, any()}

Execute the module's forward pass.

parameters(module)

(optional) 
@callback parameters(module :: t()) :: [Dspy.Parameter.t()]

Get the module's optimizable parameters.

update_parameters(module, parameters)

(optional) 
@callback update_parameters(module :: t(), parameters :: [Dspy.Parameter.t()]) :: t()

Update the module with new parameters.

Functions

compose(modules)

Compose multiple modules in sequence.

forward(module, inputs)

Execute a module's forward pass.

parallel(modules)

Run modules in parallel and merge results.

parameters(module)

Get a module's parameters if it implements the callback.

update_parameters(module, parameters)

Update a module's parameters if it implements the callback.