Interestingly this was missing from model card so I have tested quite a few different prompts so far and it also did very good with Phind one.
I have not seen it produce code in multiple chunks but sometimes I add \n```python at the end of the prompt and it fixes the issue
Pros of this model is that its consitent in the thigs it gets right (and wrong as well), while Phind is very unsure about a lot of different questions. I'm thorn which of those two is still the best codellama
Apparently, correct prompt is Alpaca:
"You are an intelligent programming assistant.
Instruction:
Implement a linked list in C++
Response:"
Interestingly this was missing from model card so I have tested quite a few different prompts so far and it also did very good with Phind one.
I have not seen it produce code in multiple chunks but sometimes I add \n```python at the end of the prompt and it fixes the issue
Pros of this model is that its consitent in the thigs it gets right (and wrong as well), while Phind is very unsure about a lot of different questions. I'm thorn which of those two is still the best codellama