llama-cpp-python
llama-cpp
wheel
windows
cuda-12
blackwell
sm_100
sm_90
sm_89
sm_86
sm_80
sm_75
sm_72
sm_70
sm_62
sm_61
cp312
Instructions to use trajis-tech/llama-cpp-python-trajis-tech-nonavx512-cuda with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use trajis-tech/llama-cpp-python-trajis-tech-nonavx512-cuda with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="trajis-tech/llama-cpp-python-trajis-tech-nonavx512-cuda", filename="{{GGUF_FILE}}", )output = llm( "Once upon a time,", max_tokens=512, echo=True ) print(output)
- Notebooks
- Google Colab
- Kaggle
metadata
license: mit
tags:
- llama-cpp
- llama-cpp-python
- wheel
- windows
- cuda-12
- blackwell
- sm_100
- sm_90
- sm_89
- sm_86
- sm_80
- sm_75
- sm_72
- sm_70
- sm_62
- sm_61
- cp312
library_name: llama-cpp-python
llama-cpp-python (Windows CUDA build)
Prebuilt wheel for:
- llama_cpp_python 0.3.16
- Windows x64
- Python 3.12 (cp312)
- CUDA enabled
- AVX512 disabled
- Supports NVIDIA 10 / 20 / 30 / 40 / 50 series GPUs
- Trajis SmartSRT 1.0.0
Install
Direct install:
Or download manually and install:
pip install llama_cpp_python-0.3.16-cp312-cp312-win_amd64.whl
Uninstall
pip uninstall llama-cpp-python
Requirements
- Windows 64-bit
- Python 3.12
- NVIDIA GPU
- CUDA Toolkit installed