Questions on creating instruction data

[henryhungle]

Thanks for the great work!

I have a few questions regarding data creation of xP3 after following the guide here to create instruction data on the code language subset.

1. I noticed the total samples of the public processed data (from here) on the code split is 2707724. However, my resulting data following the above github guide is much more than that (approximately >3M samples). I wonder if there were any additional post-processing to get the final instruction data for tuning?

2. Following the above github guide, I noticed there was no prompt for this particular dataset State Changes. I got this warning when running the creation code: Tried instantiating `DatasetTemplates` for Fraser/python-state-changes, but no prompts found. Please ignore this warning if you are creating new prompts for this dataset.

Is this dataset not assigned with any prompt (similar to how HumanEval was treated). Or is the below version of PromptSource I used is not correct: git clone -b tr13 https://github.com/Muennighoff/promptsource.git & install cd promptsource; pip install -e .

[Muennighoff]

Hey, thanks for the thorough investigation!

1. This could be due to the merging of the files. When you load from https://huggingface.co/datasets/bigscience/xP3all it loads a file called merged.jsonl for each directory, which are all individual jsonl files merged and deduplicated (https://github.com/bigscience-workshop/bigscience/blob/57086158464c4e514e8e9e3d6f77eed4865e20e4/data/xp3/xp3_jsonl_to_meg.slurm#L80).
2. Good point python-state-changes did not make it into the dataset - Not sure why. You could write some prompts for it & add it for your dataset. Your promptsource version is correct.