Inject `prompt_type` into header of current/future evaluation data
This MR includes two parts:
-
The evaluation script that is used to collect model response data is updated to include
prompt_type
(currently either "generation" or "completion") in the header. This is being added as a result of realizing the convenience of having the data field during analysis. -
The current production data is updated to have
prompt-type
included in the headers.
Validating new header
A test run of generating new prompt results to prove that the header has the prompt_type
field in the JSON object response.