请问如何在Jetpack5.1.2系统中配置适配的Cuda+pytorch+torchvision版本

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Orin
1

1、安装Jetpack6.0，Cuda12.2，配置pytorch+torchvision后，torch.cuda_is_available() 显示：False。
（1）NVIDIA-smi 显示 Driver Version: N/A；
（2）nvcc 显示正常。
2、安装Jetpack5.1，Cuda11.4，配置pytorch+torchvision后，torch.cuda_is_available() 显示：False。

3

Hi,

Could you share more about how you install PyTorch and TorchVision?

Thanks.

4

感谢回复。刷机到6.0之后，运行nvidia-smi，提示：Driver Version: N/A（正常情况应该显示已安装的驱动版本号才对）。于是我安装Cuda12.2开发工具包，当我安装 * Python 3.10 - torch-2.1.0-cp310-cp310-linux_aarch64.whl (USE_DISTRIBUTED=on)
后，通过执行：python，torch.cuda_is_available() 显示：False。也就是说，GPU根本没有得到识别。

5

另外， torch-2.2.0a0+6a974be.nv23.11-cp310-cp310-linux_aarch64.whl 对应的torchvision版本是多少？

6

附上问题的截图

012871b85fc7e9fab2f18a1b3f9d599
012871b85fc7e9fab2f18a1b3f9d5991920×1440 294 KB

由上图可知，显卡驱动并未识别，该如何修复？

7

Hi,

Based on the image above, it looks like you have two GPUs on the device.
Index 0 is Orin and the other is N/A.

Are you using Jetson AGX Orin since the device should have only one GPU?

Thanks.

8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.