However, when I use pytorch to detect the number and name of GPU, the related codes and displays are shown below
print('Number of gpu: ', torch.cuda.device_count())
print('its name: ', torch.cuda.get_device_name(0))
Number of gpu: 1
its name: NVIDIA GeForce RTX 3070 Ti
(1) Why the numbers of GPUs detected by the two methods are different?
(2) For Mujoco, how can I check the names of the detected GPUs?
(3) After I installed the Nvidia driver, the default installation path was /usr/src instead of /usr/lib. Then, I use ln -s to create a new folder /usr/lib/nvidia-510. However, there is no libGL.so mentioned in some other issues.
When I tried to render the Mujoco model with a specified GPU, I got some confusion, which may be very simple.
When the following code is executed, I can find
Found 3 GPUs for rendering. Using device 0
in the Pycharm console window.However, when I use pytorch to detect the number and name of GPU, the related codes and displays are shown below
(1) Why the numbers of GPUs detected by the two methods are different?
(2) For Mujoco, how can I check the names of the detected GPUs?
(3) After I installed the Nvidia driver, the default installation path was
/usr/src
instead of/usr/lib
. Then, I useln -s
to create a new folder/usr/lib/nvidia-510
. However, there is nolibGL.so
mentioned in some other issues.Hope to get some replies!