opto.optimizers.textgrad.TextGrad#

class TextGrad[source]#

Attributes

propagator	Return a Propagator object that can be used to propagate feedback in backward.
trace_graph	Aggregate the graphs of all the parameters.

Methods

backward(node, *args, **kwargs)	Propagate the feedback backward.
call_llm(system_prompt, user_prompt[, verbose])	Call the LLM with a prompt and return the response.
default_propagator()	Return the default Propagator object of the optimizer.
propose(*args, **kwargs)	Propose the new data of the parameters based on the feedback.
step(*args, **kwargs)	Update the parameters based on the feedback.
update(update_dict)	Update the trainable parameters given a dictionary of new data.
zero_feedback()	Reset the feedback.

__init__(parameters: List[ParameterNode], llm: AutoGenLLM | None = None, *args, propagator: Propagator | None = None, objective: None | str = None, max_tokens=4096, log=False, **kwargs)[source]#

__new__(**kwargs)#