Open sh0416 opened 1 month ago
https://github.com/SalesforceAIResearch/CodeChain/blob/85452f4009e1d8adf9fe2e7f4898b95e014c6a5e/src/processing.py#L57-L63
The code seems that the generation is filtered based on functional correctness.
Could we need it to reproduce your results?
The reason to ask is that WizardCoder could not produce any correct solution during round 0, which raises an error.
https://github.com/SalesforceAIResearch/CodeChain/blob/85452f4009e1d8adf9fe2e7f4898b95e014c6a5e/src/processing.py#L57-L63
The code seems that the generation is filtered based on functional correctness.
Could we need it to reproduce your results?