4 ms·

good to see code llama supported through Continue already. Are you seeing good results with code llama yet?

• sestinj 3y ago
  It's definitely a knowledgable model. I'm seeing a bit of trouble with stopping at the right point and managing chat context, but this is all a matter of prompt engineering. Very promising, and exciting work to be done!
  • jstummbillig 3y ago
    > I'm seeing a bit of trouble with stopping at the right point and managing chat context, but this is all a matter of prompt engineering Can you explain what you mean by this, maybe provide some examples?
    • sestinj 3y ago
      Yeah, I asked it to write bubble sort in python, and it did so perfectly, but then decided to add a couple dozen parentheses to the end. I then followed up asking it to remove these, and it didn't seem to grasp that there was a conversation, as it just rewrote the same thing. All said, this was after like 5 minutes of playing with the 13B model, and was not using any kind of human/assistant formatting, hence the need for at least simple prompting to make that work (or fine-tuning if it isn't trained yet on conversational data)