langchain/libs/experimental/langchain_experimental/tot/thought_generation.py
Nuno Campos c0d67420e5
Use a submodule for pydantic v1 compat (#9371)
<!-- Thank you for contributing to LangChain!

Replace this entire comment with:
  - Description: a description of the change, 
  - Issue: the issue # it fixes (if applicable),
  - Dependencies: any dependencies required for this change,
- Tag maintainer: for a quicker response, tag the relevant maintainer
(see below),
- Twitter handle: we announce bigger features on Twitter. If your PR
gets announced and you'd like a mention, we'll gladly shout you out!

Please make sure your PR is passing linting and testing before
submitting. Run `make format`, `make lint` and `make test` to check this
locally.

See contribution guidelines for more information on how to write/run
tests, lint, etc:

https://github.com/hwchase17/langchain/blob/master/.github/CONTRIBUTING.md

If you're adding a new integration, please include:
1. a test for the integration, preferably unit tests that do not rely on
network access,
2. an example notebook showing its use. These live is docs/extras
directory.

If no one reviews your PR within a few days, please @-mention one of
@baskaryan, @eyurtsev, @hwchase17, @rlancemartin.
 -->
2023-08-17 16:35:49 +01:00

95 lines
3.0 KiB
Python

"""
We provide two strategies for generating thoughts in the Tree of Thoughts (ToT)
framework to avoid repetition:
These strategies ensure that the language model generates diverse and
non-repeating thoughts, which are crucial for problem-solving tasks that require
exploration.
"""
from abc import abstractmethod
from typing import Any, Dict, List, Tuple
from langchain.chains.llm import LLMChain
from langchain.prompts.base import BasePromptTemplate
from langchain_experimental.pydantic_v1 import Field
from langchain_experimental.tot.prompts import COT_PROMPT, PROPOSE_PROMPT
class BaseThoughtGenerationStrategy(LLMChain):
"""
Base class for a thought generation strategy.
"""
c: int = 3
"""The number of children thoughts to propose at each step."""
@abstractmethod
def next_thought(
self,
problem_description: str,
thoughts_path: Tuple[str, ...] = (),
**kwargs: Any
) -> str:
"""
Generate the next thought given the problem description and the thoughts
generated so far.
"""
class SampleCoTStrategy(BaseThoughtGenerationStrategy):
"""
Sample thoughts from a Chain-of-Thought (CoT) prompt.
This strategy works better when the thought space is rich, such as when each
thought is a paragraph. Independent and identically distributed samples
lead to diversity, which helps to avoid repetition.
"""
prompt: BasePromptTemplate = COT_PROMPT
def next_thought(
self,
problem_description: str,
thoughts_path: Tuple[str, ...] = (),
**kwargs: Any
) -> str:
response_text = self.predict_and_parse(
problem_description=problem_description, thoughts=thoughts_path, **kwargs
)
return response_text if isinstance(response_text, str) else ""
class ProposePromptStrategy(BaseThoughtGenerationStrategy):
"""
Propose thoughts sequentially using a "propose prompt".
This strategy works better when the thought space is more constrained, such
as when each thought is just a word or a line. Proposing different thoughts
in the same prompt completion helps to avoid duplication.
"""
prompt: BasePromptTemplate = PROPOSE_PROMPT
tot_memory: Dict[Tuple[str, ...], List[str]] = Field(default_factory=dict)
def next_thought(
self,
problem_description: str,
thoughts_path: Tuple[str, ...] = (),
**kwargs: Any
) -> str:
if thoughts_path not in self.tot_memory or not self.tot_memory[thoughts_path]:
new_thoughts = self.predict_and_parse(
problem_description=problem_description,
thoughts=thoughts_path,
n=self.c,
**kwargs
)
if not new_thoughts:
return ""
if isinstance(new_thoughts, list):
self.tot_memory[thoughts_path] = new_thoughts[::-1]
else:
return ""
return self.tot_memory[thoughts_path].pop()