Yuxiang Wei

Yeah we are doing some relevant follow-ups. Stay tuned for future updates:)

Thanks for asking. I'm not very familiar with ooba. Would you mind providing more context?

Here is the normal text-generation prompt for Magicoder: ``` MAGICODER_PROMPT = """You are an exceptionally intelligent coding assistant that consistently delivers accurate and reliable responses to user instructions. @@ Instruction...

I have the same issue after inspecting the data downloaded from http://eaidata.bmk.sh/data/github_small.jsonl.zst. It seems the value of the 'file_name' key is identical for every repo.

Hi @mstallone, thanks for asking. Let me attach the logs for each training step here. To best reproduce the HumanEval score, you can follow the steps outlined in the `evaluation`...

Ah I think it's due to the dropout. starcoder2-15b sets the dropout value to `0.1` by default. We did not apply dropout during finetuning, so that's why your loss is...

I don't think I enabled gradient checkpointing. This set of hyperparameters just barely allows the starcoder2-15b to fit on one A100 80G. If it helps, here is the JSON form...

Hi, StarCoder-15b is a base model, and we use it to generate data and create starcoder2-instruct. The main pipeline code is here: https://github.com/bigcode-project/selfcodealign/blob/46749cceeb84cf1773deb5c157b04930acebc5cd/src/star_align/self_ossinstruct.py.

You can find the instructions from an earlier selfcodealign, [starcoder2-instruct](https://github.com/bigcode-project/selfcodealign/tree/starcoder2-instruct). Sorry the instructions for the new version has not been updated yet. It's mostly compatible with the old one, but...

Thanks @jacobkahn. I think it's pretty detailed! The TL;DR is that even without internet access, the agent can still cheat by peeking at future commits using `git`, because such information...