Training data set
I thought I read somewhere that this dataset (https://github.com/BibleNLP/ebible) was used to train this collection of models. Now I can't find that reference. I only see this dataset (https://huggingface.co/datasets/sleepdeprived3/Reformed-Christian-Bible-Expert). Am I mistaken about the BibleNLP/ebible ?
I never saw that BibleNLP/ebible one. I make my own datasets such as the Reformed-Christian-Bible-Expert one you found. I have a newer version available now I'll try to remember to upload it when I can get around to it.
Would you be willing to collaborate on a process for specializing a Mistral 7B language model on this dataset (https://github.com/BibleNLP/ebible) using the QLoRA technique? I have a detailed approach for doing this, but I've never done any training or fine-tuning.
I don't see a dataset available there. A dataset would be user/assistant turns.