Chunking functions for all languages
Summary
We need to support all 12 languages when building testcases using different methods. Current there are 2 methods for building the raw prompt:
We define this metric as Opportunity
Opportunity is of two types for now:
- current block: the smallest block where the cursor belongs, this is usually a function.
- current context: the top level entity that where the cursor belongs, this can be a class, a module etc.
In both methods, the expected completion is everything below the cursor within the smallest block. We are trying to evaluate the model's ability to complete the current block rather than to complete the rest of the file.
Language Supports Status:
-
python -
javascript -
typescript -
c: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/29 -
cpp -
csharp: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/29 -
go: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/29 -
java -
php -
ruby: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/29 -
rust: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/merge_requests/29 -
scala
Based on the existing languages in content_v1 BQ table,
Iteration 1: Python, Javascript , Typescript
Iteration 2: c, csharp, go, ruby, rust
16.4 Iteration 3: Rest of the languages
Edited by Mon Ray