Open gsireesh opened 5 months ago
The current iteration of the GPT-3.5 predictor currently returns invalid json more often than not for many document sections. Figure out why, and how to modify the prompt for more reliable results.
Fixed GPT_Predictor.py with a different API call method
Also added the test script for future debugging use
Created a side-branch "GPT_API_Fix" and pushed it to Github repo
The current iteration of the GPT-3.5 predictor currently returns invalid json more often than not for many document sections. Figure out why, and how to modify the prompt for more reliable results.