Update explanations to be consistent with paper

Upload tokenizer

Update model to v0.2 fine-tuned on mistral_7b_v01 which is consistent with the paper.

initial commit