Gpt math solver #991
Gpt math solver #991
Conversation
@microsoft-github-policy-service agree |
flaml/autogen/oai/completion.py
Outdated
| # try to format the message with the data instance | ||
| content = m["content"].format(**data_instance) | ||
| except Exception: | ||
| # if it fails, use the raw message |
There was a problem hiding this comment.
Add a warning message here to remind that the program failed to format the message with the data instance and thus raw message is being used?
There was a problem hiding this comment.
There might be a long sequence of texts and will be a lot of warning. We should assume that when use inputs messages instead of prompt, these message are less likely containing formatted strings. We can add comments to the doc to explain.
|
Thanks for the contribution. Could you please run pre-commit first: https://microsoft.github.io/FLAML/docs/Contribute#pre-commit to ensure the code passes the "Code formatting" test? |
|
Add a test case? |
|
Is this PR intended for merging? If not, please make it a draft. |
|
Please resolve the conflict. |
| 'model': model, | ||
| "max_tokens": max_tokens, | ||
| 'messages' : [ | ||
| {"role": "system", "content": "You are a helpful assistant."}, |
There was a problem hiding this comment.
What about making the prompt before the "Problem:" a system message?
flaml/autogen/math_solver/main.py
Outdated
| results = selfconsistency.vanilla_voting(responses["responses"], problem['solution']) | ||
| print(results['success_vote'], results['votes']) | ||
|
|
||
| def tool_voting_one_category(model, problem_set, saving_folder, n=2, n_per_time=1): |
Co-authored-by: Chi Wang <[email protected]>
Co-authored-by: Chi Wang <[email protected]>
…into gpt_math_solver
Why are these changes needed?
Related issue number
Closes #990
Checks