rccmsu/ruadapt_saiga2_7b_v0.1

Use in the same way as IlyaGusev/saiga27blora.

WARNING! Load tokenizer as AutoTokenizer.frompretrained(modelpath, use_fast=True)

Up to 60% faster generation and 35% training (on identical russian text sequences!) with HF because of different tokenizer.

Colab: https://colab.research.google.com/drive/109ZhEB6STy-0jO-Z4ttkWr1jgFCTRW?usp=sharing

Paper: Tikhomirov M., Chernyshev D. Impact of Tokenization on LLaMa Russian Adaptation //arXiv preprint arXiv:2312.02598. – 2023.

Model description

Instruction version (Saiga datasets) of Russian adaptation of LLaMa-2-7B by replacing the tokenizer. Paper: Tikhomirov M.M., Chernyshev D.I., Impact of Tokenization on LLaMa Russian Adaptation (will be soon)