Hi, @bordernone , can you please create a new column in the video table named ada_search? It needs to store a NumPy array for each row. IDK what would be the best format in SQL for storing that? To understand what that looks like and how it is utilized by the new dialogue manager, please go to my amc-dev-dm branch and the dm-dev notebook: https://github.com/CAMeL-Lab/TOIA-2.0/blob/amc-dev-dm/server/toia-dm/notebooks/dm-dev.ipynb.
It would help if you run until the 'get data' section is included, then you can jump to the end to 'using OpenAI GPT-3'.
We need embeddings precomputed and stored into the DB for the new DM to work. I created a separate script to update the table when we have the new column. At some point, we shall also consider saving the embedding straight when the user preserves or edits a video.
AMChierici
commented
2 years ago
Hi, @bordernone , can you please create a new column in the video table named
ada_search? It needs to store a NumPy array for each row. IDK what would be the best format in SQL for storing that? To understand what that looks like and how it is utilized by the new dialogue manager, please go to myamc-dev-dmbranch and the dm-dev notebook: https://github.com/CAMeL-Lab/TOIA-2.0/blob/amc-dev-dm/server/toia-dm/notebooks/dm-dev.ipynb.It would help if you run until the 'get data' section is included, then you can jump to the end to 'using OpenAI GPT-3'.
We need embeddings precomputed and stored into the DB for the new DM to work. I created a separate script to update the table when we have the new column. At some point, we shall also consider saving the embedding straight when the user preserves or edits a video.
Let me know if you have any questions!