Hello. I’m starting out in my HuggingFace journey with the intent of using an LLM to expedite feature development in a codebase.
I understand that StarCoder seems to be the way to go and my idea is to fine-tune it on my code. But the page explicitly states the following:
As such it is not an instruction model and commands like “Write a function that computes the square root.” do not work well. However, by using the Tech Assistant prompt you can turn it into a capable technical assistant.
The Tech Assistant prompt appears to be a training set of 650 records. This seems way too small to be sufficiently versatile. Is this the recommended combination of tools for my purpose (to build an LLM which I can ask to generate new code) or are there other approaches in the HuggingFace ecosystem?