5

I have installed the pytorch, and would like to check are there any script to test whether the installation is correct, e.g., whether it can enable CUDA or not, etc?

user288609
  • 12,465
  • 26
  • 85
  • 127
  • Possible duplicate of [How to check if pytorch is using the GPU?](https://stackoverflow.com/questions/48152674/how-to-check-if-pytorch-is-using-the-gpu) – prosti Sep 18 '19 at 11:30

4 Answers4

5

Coming to your 1st question, In your python script.... just add

import torch

if this gives "ModuleNotFoundError: No module named 'torch'", then your pytorch installation is not complete

And your 2nd question to check if your pytorch is using cuda,use this

torch.cuda.is_available()

this will return True if your pytorch is using cuda

Prajot Kuvalekar
  • 5,128
  • 3
  • 21
  • 32
  • This alone will not verify that pytorch will work. See [this issue](https://github.com/pytorch/pytorch/issues/31285) – Pro Q May 09 '22 at 07:58
  • there are some issues with TITAN series – Prajot Kuvalekar May 09 '22 at 09:50
  • Given the comments on [the issue](https://github.com/pytorch/pytorch/issues/31285), and my own personal case, it seems that for any GPU that has `sm_86` capabilities, this script will run successfully, but will end up failing when you try to run real code/do ML training. [Bai's answer](https://stackoverflow.com/a/71988947/5049813), which includes more tests of the true functionality, will fail if attempted on these sorts of GPUs. – Pro Q May 09 '22 at 10:29
4

You can use the collect_env.py script provided in the PyTorch utils folder.

Its output is as follows:

Collecting environment information...
PyTorch version: 1.2.0
Is debug build: No
CUDA used to build PyTorch: 10.0.130

OS: Ubuntu 16.04.6 LTS
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.11) 5.4.0 20160609
CMake version: version 3.14.6

Python version: 3.7
Is CUDA available: Yes
CUDA runtime version: Could not collect
GPU models and configuration: GPU 0: GeForce RTX 2080
Nvidia driver version: 410.48
cuDNN version: /usr/lib/x86_64-linux-gnu/libcudnn.so.7.4.1

Versions of relevant libraries:
[pip] numpy==1.16.4
[pip] torch==1.2.0
[pip] torchsample==0.1.3
[pip] torchsummary==1.5.1
[pip] torchvision==0.4.0a0+6b959ee
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      243  
[conda] mkl-service               2.0.2            py37h7b6447c_0  
[conda] mkl_fft                   1.0.14           py37ha843d7b_0  
[conda] mkl_random                1.0.2            py37hd81dba3_0  
[conda] pytorch                   1.2.0           py3.7_cuda10.0.130_cudnn7.6.2_0    pytorch
[conda] torchsample               0.1.3                    pypi_0    pypi
[conda] torchsummary              1.5.1                    pypi_0    pypi
[conda] torchvision               0.4.0                py37_cu100    pytorch
kHarshit
  • 11,362
  • 10
  • 52
  • 71
3

If you installed it from here you are doing fine.

enter image description here

Check this:

import torch
dev = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
print(dev)

If you have your GPU installed correctly you should have nvidia-smi. (On Windows it should be inside C:\Program Files\NVIDIA Corporation\NVSMI)

prosti
  • 42,291
  • 14
  • 186
  • 151
1

You can use my code. This repo implements a set of PyTorch environment checker and cuda-based operators, which helps you verify whether your GPU-based PyTorch is installed properly.

https://github.com/BAI-Yeqi/PyTorch-Verification

Jake2099
  • 11
  • 1