Autocomplete: Tweak StarCoder temperature #966
Merged
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Outdated suggestions cannot be applied.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.
This changes the default temperature for StarCoder from
0.4
to0.2
which is also used in their own testing. In my experiments, this made the model behave a bit more predictable while still providing enough variation.CONTROVERSIAL: I also tweaked the example Bee added in #945. I’m happy to revert though. StarCoder was not completing this and I think it was because there are two reference function in that file that are implemented the same way with the same line missing so starcoder did pattern match and thought that this is the expected result. I feel that the example is a bit contrived so I changed it into something that I would encounter while developing.
Test plan
Ran the completions test suite and made sure the output looked good:
cd vscode
pnpm run generate:completions