| 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](https://huggingface.co/IlyaGusev/saiga2_7b_lora). | |
| 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](https://raw.githubusercontent.com/IlyaGusev/rulm/master/self_instruct/src/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 | |