NousResearch

/

Yarn-Mistral-7b-128k

like 567

I would like to know if your Mistral code can perform pretraining from scratch, and if so, what kind of script would be needed? For example, something like train.py?