—microsoft one, very useful
Langchain is meant to be able to do this using their application of ReAct or chain of thought.
Regex + CFGs
How to fine-tune a model to be in instruct?
- Provide the LLM with some examples of
instructionsand their corresponding outputs.
- Consider the quality and quantity of the data, the learning rate and regularization,
the freezing and unfreezing of layers,and the evaluation and validation