ModelLaboratory#

class langchain.model_laboratory.ModelLaboratory(chains: Sequence[Chain], names: List[str] | None = None)[source]#

A utility to experiment with and compare the performance of different models.

Initialize the ModelLaboratory with chains to experiment with.

Parameters:

  • chains (Sequence[Chain]) – A sequence of chains to experiment with.

  • variable. (Each chain must have exactly one input and one output)

  • names (Optional[List[str]])

names (Optional[List[str]]): Optional list of names corresponding to each chain.

If provided, its length must match the number of chains.

Raises:

  • ValueError – If any chain is not an instance of Chain.

  • ValueError – If a chain does not have exactly one input variable.

  • ValueError – If a chain does not have exactly one output variable.

  • ValueError – If the length of names does not match the number of chains.

Parameters:

  • chains (Sequence[Chain])

  • names (Optional[List[str]])

Methods

__init__(chains[, names])

Initialize the ModelLaboratory with chains to experiment with.

compare(text)

Compare model outputs on an input text.

from_llms(llms[, prompt])

Initialize the ModelLaboratory with LLMs and an optional prompt.
__init__(chains: Sequence[Chain], names: List[str] | None = None)[source]#

Initialize the ModelLaboratory with chains to experiment with.

Parameters:

  • chains (Sequence[Chain]) – A sequence of chains to experiment with.

  • variable. (Each chain must have exactly one input and one output)

  • names (List[str] | None)

names (Optional[List[str]]): Optional list of names corresponding to each chain.

If provided, its length must match the number of chains.

Raises:

  • ValueError – If any chain is not an instance of Chain.

  • ValueError – If a chain does not have exactly one input variable.

  • ValueError – If a chain does not have exactly one output variable.

  • ValueError – If the length of names does not match the number of chains.

Parameters:

  • chains (Sequence[Chain])

  • names (List[str] | None)

compare(text: str) → None[source]#

Compare model outputs on an input text.

If a prompt was provided with starting the laboratory, then this text will be fed into the prompt. If no prompt was provided, then the input text is the entire prompt.

Parameters:

text (str) – input text to run all models on.

Return type:

None

classmethod from_llms(llms: List[BaseLLM], prompt: PromptTemplate | None = None) → ModelLaboratory[source]#

Initialize the ModelLaboratory with LLMs and an optional prompt.

Parameters:

  • llms (List[BaseLLM]) – A list of LLMs to experiment with.

  • prompt (Optional[PromptTemplate]) – An optional prompt to use with the LLMs. If provided, the prompt must contain exactly one input variable.

Returns:

An instance of ModelLaboratory initialized with LLMs.

Return type:

ModelLaboratory

Examples using ModelLaboratory