Fix huggingface response caching bug #659
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅ |
rogthefrog
approved these changes
Nov 1, 2024
| ) -> SUTResponse: | ||
| completions = [] | ||
| for choice in response.choices: | ||
| text = choice.message.content | ||
| text = choice["message"]["content"] |
There was a problem hiding this comment.
You may want to try the benedict library. It wraps dict and gives you convenient functionality like accessing elements via dot notation or dict notation interchangeably (choice["message"] or choice.message).
wpietri
approved these changes
Nov 1, 2024
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The huggingface chat_completions API returns a dataclass/dict object
ChatCompletionOutput. It inherits fromdict, which is a "typeable" data structure and thus allows it to be cached. However, there was a bug when running this SUT on a prompt that has been previously cached. Because it's a dataclass and not a plain dict, the deserialization produced an error.This PR addresses this bug by putting the response object into pydantic form before returning it in evaluate.
I also added some testing to make sure all SUTs are cacheable.