Skip to content

Ask the LLM to return a patch instead of full optimized code? #42

Description

@Xia-Weiwen

It may exceed the output length limitation of the LLM if the full optimized code is requested.
The message I saw:

<incomplete triton code>
Expected to find output fields in the LM response: [optimized_code, next_thought]
Actual output fields parsed from the LM response: [optimized_code]

How about asking LLM to only return a patch instead of the full code?

A straightforward idea is to modify the prompt to the LLM to tell it to return a patch instead of the full optimized code. But the change may also affect the parser and other operations after LLM returns. Can anyone help with it? Thanks.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions