metadata
datasets:
- IlyaGusev/ru_turbo_alpaca
- IlyaGusev/ru_turbo_saiga
- IlyaGusev/ru_sharegpt_cleaned
- IlyaGusev/oasst1_ru_main_branch
- IlyaGusev/ru_turbo_alpaca_evol_instruct
- lksy/ru_instruct_gpt4
language:
- ru
inference: false
pipeline_tag: conversational
license: llama2
Llama.cpp compatible versions of an original 7B model.
Download one of the versions, for example model-q4_K.gguf.
wget https://huggingface.co/IlyaGusev/saiga2_7b_gguf/resolve/main/model-q4_K.gguf
Download interact_llamacpp.py
wget https://raw.githubusercontent.com/IlyaGusev/rulm/master/self_instruct/src/interact_llamacpp.py
How to run:
pip install llama-cpp-python fire
python3 interact_llamacpp.py model-q4_K.gguf
System requirements:
- 10GB RAM for q8_0 and less for smaller quantizations