Edit model card

Descripton:

This is ruadapt version of upstage/SOLAR-10.7B-v1.0 model with tokenizer replacement. Additionally to previous work, the model was adapted in two stages: 1) vocabulary optimization, and 2) additional attention fine-tuning using LoRa.

How to cite:

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

Downloads last month
745
Safetensors
Model size
10.7B params
Tensor type
FP16
·
Inference API
Model is too large to load in Inference API (serverless). To try the model, launch it on Inference Endpoints (dedicated) instead.