安装显卡驱动

https://www.nvidia.cn/drivers/details/242443/

安装最新的驱动，不能用windows自动安装的驱动，版本太低了

安装CUDA

https://developer.nvidia.com/cuda-12-6-3-download-archive?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local

目前相关库支持的最新版本是12.6，不要装最新的。

显卡算力表：

算力与CUDA版本对应表：

下载对应版本的CUDA

安装cuDNN

https://developer.nvidia.com/cudnn-downloads?target_os=Windows&target_arch=x86_64&target_version=Agnostic&cuda_version=12

cuDNN（CUDA Deep Neural Network library）是专为深度学习优化的GPU加速库，基于NVIDIA的CUDA平台构建，依赖CUDA提供的底层并行计算能力，专注于高效实现神经网络操作（如卷积、池化），两者关系类似于“操作系统与应用软件”或“工作台与专用工具”。‌

版本要和CUDA的对应

安装pytorch

https://pytorch.org/get-started/locally/

选择好后，运行最下面的命令

确认pythorch使用的是显卡：

在ipython中运行：

In [1]: import torch

In [2]: print(torch.__version__)
2.6.0+cu126

"cu126"说明使用的是显卡。

如果是“cpu"说明用的是cpu。

安装bitsandbytes

pip install --upgrade bitsandbytes[cuda126]

“126”是cuda一样的版本号。

查看安装是否成功：

python -m bitsandbytes

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++ BUG REPORT INFORMATION ++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++++++ OTHER +++++++++++++++++++++++++++
CUDA specs: CUDASpecs(highest_compute_capability=(8, 6), cuda_version_string='126', cuda_version_tuple=(12, 6))
PyTorch settings found: CUDA_VERSION=126, Highest Compute Capability: (8, 6).
To manually override the PyTorch CUDA version please see: https://github.com/TimDettmers/bitsandbytes/blob/main/docs/source/nonpytorchcuda.mdx
The directory listed in your path is found to be non-existent: \XGW-AI
The directory listed in your path is found to be non-existent: C:\Users\xgw10\Documents\WindowsPowerShell\Modules
CUDA SETUP: WARNING! CUDA runtime files not found in any environmental path.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++++++++++++++++++++ DEBUG INFO END ++++++++++++++++++++++
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Checking that the library is importable and CUDA is callable...
SUCCESS!
Installation was successful!