PyTorch 和 CUDA 驱动程序

我安装了 CUDA 9.2。例如：

(base) c:\>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Wed_Apr_11_23:16:30_Central_Daylight_Time_2018
Cuda compilation tools, release 9.2, V9.2.88

我使用以下命令在 Windows 10 上安装了 PyTorch：

conda install pytorch cuda92 -c pytorch
pip3 install torchvision

我运行了测试脚本：

(base) c:\>python
Python 3.6.5 |Anaconda custom (64-bit)| (default, Mar 29 2018, 13:32:41) [MSC v.1900 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> from __future__ import print_function
>>> import torch
>>> x = torch.rand(5, 3)
>>> print(x)
tensor([[0.7041, 0.5685, 0.4036],
        [0.3089, 0.5286, 0.3245],
        [0.3504, 0.8638, 0.1118],
        [0.6517, 0.9209, 0.6801],
        [0.0315, 0.1923, 0.8720]])
>>> quit()

所以对于，很好。然后我跑了：

(base) c:\>python
Python 3.6.5 |Anaconda custom (64-bit)| (default, Mar 29 2018, 13:32:41) [MSC v.1900 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
False
>>>

为什么 PyTorch 说 CUDA 不可用？

GPU为计算能力3.0 Quadro K3000M：

(base) C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi.exe 

Mon Oct 01 16:36:47 2018 
NVIDIA-SMI 385.54 Driver Version: 385.54 
-------------------------------+----------------------+---------------------- 
GPU Name TCC/WDDM              | Bus-Id Disp.A        | Volatile Uncorr. 
ECC Fan Temp Perf Pwr:Usage/Cap| Memory-Usage         | GPU-Util Compute M. 
0 Quadro K3000M WDDM           | 00000000:01:00.0 Off | 
N/A N/A 35C  P0 N/A / N/A      | 29MiB / 2048MiB      | 0% Default 

自从https://github.com/pytorch/pytorch/releases/tag/v0.3.1，PyTorch 二进制版本删除了对具有 CUDA 功能 3.0 的旧 GPU 的支持。根据https://en.wikipedia.org/wiki/CUDA，Quadro K3000M的计算能力为3.0。

因此，您可能必须从源代码构建 pytorch 或尝试其他软件包。请参阅此主题以获取更多信息——https://discuss.pytorch.org/t/pytorch-no-longer-supports-this-gpu-because-it-is-too-old/13803.