AssertionError: Torch not compiled with CUDA enabled

bugcreater · July 5, 2017, 2:17pm

when I ran my code in server with GPU is ok
but when I ran the code in another pc using pycharm remote local,it shows:
“AssertionError: Torch not compiled with CUDA enabled”

albanD · July 5, 2017, 2:59pm

If there is no GPU on the machine and you try to use cuda elements, it will fail with this error message because you try to use cuda tools but they are not installed.

bugcreater · July 5, 2017, 3:02pm

Is there other method can I use GPU?Jupyter?

albanD · July 5, 2017, 3:06pm

Is your pycharm suppose to run on the server with GPU?
It doesn’t matter what python interpreter you are using. To be able to use GPU, you need a computer with a GPU and install pytorch on this computer.

ati · October 16, 2019, 2:54pm

Hi…i have the same problem
i installed pytorch and my laptop has GPU also
what’s wrong with that that i take the same error?

albanD · October 16, 2019, 3:09pm

Hi,

You would need to give more informations here.
Like what is your local machine, how do you install pytorch, what is already installed on the machine (do you already have cuda installed)?

ati · October 16, 2019, 3:57pm

I am running this code:

github.com

Hananel-Hazan/bindsnet/blob/master/examples/mnist/reservoir.py

import os
import numpy as np
import torch
import torch.nn as nn
import argparse
import matplotlib.pyplot as plt

from torchvision import transforms
from tqdm import tqdm

from bindsnet.analysis.plotting import (
    plot_input,
    plot_spikes,
    plot_voltages,
    plot_weights,
)
from bindsnet.datasets import MNIST
from bindsnet.encoding import PoissonEncoder
from bindsnet.network import Network
from bindsnet.network.nodes import Input

This file has been truncated. show original

I did not install Cuda and i don’t want to do
and just installed pytorch on windows10
is there any way in this code that i run it on cpu???

ati · October 16, 2019, 4:01pm

By the way , i installed torch by
pip3 install torch==1.3.0+cpu torchvision==0.4.1+cpu -f https://download.pytorch.org/whl/torch_stable.html

albanD · October 16, 2019, 5:27pm

If you don’t want to use cuda, don’t use it
Make sure that the code you run does not use cuda stuff and you won’t see this error anymore !

ati · October 16, 2019, 5:49pm

Thanks for your response @alban …yeah i don’t want it
I deleted all of cuda stuff
but i don’t know where i am wrong that still it gives me this error
in one line i see this code:
parser.add_argument("–device_id", type=int, default=0)
that i think it uses cuda as default …
do u have any idea what value i set for default that it uses cpu
i used -1 but it gives me error that it can not be negative

albanD · October 16, 2019, 5:55pm

Hi,

You should get a stack trace with the error. Can you share it? It should point at the place where cuda is being used

ati · October 16, 2019, 6:16pm

this is the full of error

type or pWarning (from warnings module):
  File "C:\python\Python37\lib\site-packages\sklearn\externals\joblib\externals\cloudpickle\cloudpickle.py", line 47
    import imp
DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
Traceback (most recent call last):
  File "C:\Users\Ati\Desktop\bindsnet-master\examples\mnist\reservoir.py", line 95, in <module>
    network.to("cuda")
  File "C:\python\Python37\lib\site-packages\torch\nn\modules\module.py", line 426, in to
    return self._apply(convert)
  File "C:\python\Python37\lib\site-packages\torch\nn\modules\module.py", line 202, in _apply
    module._apply(fn)
  File "C:\python\Python37\lib\site-packages\torch\nn\modules\module.py", line 245, in _apply
    self._buffers[key] = fn(buf)
  File "C:\python\Python37\lib\site-packages\torch\nn\modules\module.py", line 424, in convert
    return t.to(device, dtype if t.is_floating_point() else None, non_blocking)
  File "C:\python\Python37\lib\site-packages\torch\cuda\__init__.py", line 192, in _lazy_init
    _check_driver()
  File "C:\python\Python37\lib\site-packages\torch\cuda\__init__.py", line 95, in _check_driver
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabledaste code here

albanD · October 16, 2019, 7:57pm

There is the line network.to("cuda") at line 95 of your reservoir.py file.
You need to remove this .to() or change the argument to "cpu" if you don’t want to use cuda.

ati · October 16, 2019, 8:36pm

if gpu:
    network.to("cuda")

i changed it but it doesn’t work
i think it says if gpu is true then run on cuda
so i think i have to change where gpu is true and make it default as false
but i don’t know where is that

ati · October 16, 2019, 8:44pm

Hey…thank you i solved it finally by this your sentence help:
“Make sure that the code you run does not use cuda stuff”
I deleted all of Cuda , gpu and device_id words
and it’s working now
thank you

FRahmani368 · July 10, 2020, 7:43am

Hi, I have an issue with using CUDA. I got this error “torch not compiled with cuda enabled” when I wanted to define a tensor on GPU.
for pytorch installation, I used:
conda install pytorch torchvision cudatoolkit=10.2 -c pytorch
and I have Geforce GTX 1080 Ti
Nvidia CUDA driver is 11.0.197. GPU version 451.48
I am using pycharm.
when I run torch.cuda.is_available() --> I get False.
What is the problem? Please advise me. thank you

ptrblck · July 11, 2020, 11:55pm

Maybe the installation of the binary has failed.
Could you create a new conda environment, reinstall the binary, and post the install log here?

surendra_mehra · July 29, 2021, 10:47am

I have also same error.
I have install conda install pytorch torchvision torchaudio cudatoolkit=11.0 -c pytorch cuda toolkit but i have print the following command and show the result print(torch.cuda.is_available())
False.
I have created new environment and installed it.
please give me suggestion. how to resolv this error

ptrblck · July 30, 2021, 4:30am

Could you post the install log here, please?