Adding text-bison/code-bison to promptlibrary
This MR adds text-bison
and code-bison
to the ensemble of models in the prompt library for Code Suggestions. This MR refactors some of the current eval pipeline to allow for each model call and combining the data to be one table before being scored. The current implementation will not be sustainable long term, but is an intermediate solution while a more robust solution is configured.
For more details on this effort, reference this issue: https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/prompt-library/-/issues/14+.
cc @mray2020
Edited by Dylan Bernardi