Cannot install torch-GPU in my torch installation

No_Yeah · June 14, 2023, 8:37am

Hi I am trying to install new graphic card to PC but struggling to do it

Here is the graphic card NVIDIA Geforce Xp

cpuz

seems well recognized in the PC

in NVIDIA site it says its compute capability is 6.1
from official (https://developer.nvidia.com/cuda-gpus)

so in the below site, I have downloaded CUDA toolkit 10.0
from official (https://developer.nvidia.com/cuda-toolkit-archive)
with compatible Visual Studio 2017

and download cuDNN 7.3.0 to meet CUDA toolkit version
from official
unzip it into CUDA\v10.0

and added following paths in system env variables
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\extras\CUPTI\libx64
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include

then I have checked it on command prompt

 C:\Windows\system32>nvidia-smi
Wed Jun 14 17:06:13 2023
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.98                 Driver Version: 535.98       CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                     TCC/WDDM  | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA TITAN Xp              WDDM  | 00000000:83:00.0 Off |                  N/A |
| 23%   28C    P8              10W / 250W |     65MiB / 12288MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|    0   N/A  N/A      8640    C+G   C:\Windows\explorer.exe                   N/A      |
|    0   N/A  N/A     10252    C+G   ...CBS_cw5n1h2txyewy\TextInputHost.exe    N/A      |
|    0   N/A  N/A     13368    C+G   ....Search_cw5n1h2txyewy\SearchApp.exe    N/A      |
|    0   N/A  N/A     15952    C+G   ...ekyb3d8bbwe\PhoneExperienceHost.exe    N/A      |
|    0   N/A  N/A     16624    C+G   ...cal\Microsoft\OneDrive\OneDrive.exe    N/A      |
|    0   N/A  N/A     17464    C+G   ...siveControlPanel\SystemSettings.exe    N/A      |
+---------------------------------------------------------------------------------------+

C:\Windows\system32>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:04_Central_Daylight_Time_2018
Cuda compilation tools, release 10.0, V10.0.130

then I installed anaconda

and make enviroment for Deep learning setting

conda create --name conda-forge python=3.6
conda activate conda-forge

From offical pytorech site

for torch install CUDA 10.0, the command is

conda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=10.0 -c pytorch

I have used that code but got below error message

PackagesNotFoundError: The following packages are not available from current channels:

  - torchvision==0.4.0
  - pytorch==1.2.0

Current channels:

  - https://conda.anaconda.org/pytorch/win-64
  - https://conda.anaconda.org/pytorch/noarch
  - https://repo.anaconda.com/pkgs/main/win-64
  - https://repo.anaconda.com/pkgs/main/noarch
  - https://repo.anaconda.com/pkgs/r/win-64
  - https://repo.anaconda.com/pkgs/r/noarch
  - https://repo.anaconda.com/pkgs/msys2/win-64
  - https://repo.anaconda.com/pkgs/msys2/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.

so I used the below code

conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

but it only uses CPU only

The following NEW packages will be INSTALLED:

  blas               pkgs/main/win-64::blas-1.0-mkl
  cpuonly            pytorch/noarch::cpuonly-2.0-0
  dataclasses        pkgs/main/noarch::dataclasses-0.8-pyh4f3eec9_6
  freetype           pkgs/main/win-64::freetype-2.12.1-ha860e81_0
  intel-openmp       pkgs/main/win-64::intel-openmp-2023.1.0-h59b6b97_46319
  jpeg               pkgs/main/win-64::jpeg-9e-h2bbff1b_1
  lerc               pkgs/main/win-64::lerc-3.0-hd77b12b_0
  libdeflate         pkgs/main/win-64::libdeflate-1.17-h2bbff1b_0
  libpng             pkgs/main/win-64::libpng-1.6.39-h8cc25b3_0
  libtiff            pkgs/main/win-64::libtiff-4.5.0-h6c2663c_2
  libuv              pkgs/main/win-64::libuv-1.44.2-h2bbff1b_0
  lz4-c              pkgs/main/win-64::lz4-c-1.9.4-h2bbff1b_0
  mkl                pkgs/main/win-64::mkl-2020.2-256
  mkl-service        pkgs/main/win-64::mkl-service-2.3.0-py36h196d8e1_0
  mkl_fft            pkgs/main/win-64::mkl_fft-1.3.0-py36h46781fe_0
  mkl_random         pkgs/main/win-64::mkl_random-1.1.1-py36h47e9c7a_0
  numpy              pkgs/main/win-64::numpy-1.19.2-py36hadc3359_0
  numpy-base         pkgs/main/win-64::numpy-base-1.19.2-py36ha3acd2a_0
  olefile            pkgs/main/win-64::olefile-0.46-py36_0
  pillow             pkgs/main/win-64::pillow-8.3.1-py36h4fa10fc_0
  pytorch            pytorch/win-64::pytorch-1.10.2-py3.6_cpu_0
  pytorch-mutex      pytorch/noarch::pytorch-mutex-1.0-cpu
  six                pkgs/main/noarch::six-1.16.0-pyhd3eb1b0_1
  tk                 pkgs/main/win-64::tk-8.6.12-h2bbff1b_0
  torchvision        pytorch/win-64::torchvision-0.11.3-py36_cpu
  typing_extensions  pkgs/main/noarch::typing_extensions-4.1.1-pyh06a4308_0
  xz                 pkgs/main/win-64::xz-5.4.2-h8cc25b3_0
  zlib               pkgs/main/win-64::zlib-1.2.13-h8cc25b3_0
  zstd               pkgs/main/win-64::zstd-1.5.5-hd43e919_0

I have tested it in python environment and rebooted and tested… nothing works…

I am doing this for a day and this driving me so crazy please help me

ptrblck · June 14, 2023, 12:23pm

Neither your locally installed CUDA toolkit nor cuDNN will be used unless you build from source or a custom CUDA extension.
The PyTorch binaries ship with their own CUDA dependencies so try to install a newer release with a newer CUDA runtime by selecting the desired setup from our website.

No_Yeah · June 15, 2023, 5:21am

tried 11.2 but not work…

ptrblck · June 15, 2023, 5:23am

The current releases do not ship with CUDA 11.2 so it’s unclear what exactly you have installed and what exactly does not work.

No_Yeah · June 16, 2023, 12:30am

My GPU NVIDIA GeForce Tiatan Xp can support cuda version range of CUDA 10.0 ~ 11.2 so I have installed 10.0, 11.2

for 10.0 as the pytorch offical says

# CUDA 10.0
conda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=10.0 -c pytorch

so I typed it but got the following message

PackagesNotFoundError: The following packages are not available from current channels:

pytorch==1.2.0

torchvision==0.4.0

using the below

# CUDA 10.0
conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

It works but the torch, torchvision only install CPU version not GPU version

same for the CUDA 11.2

ptrblck · June 16, 2023, 1:35am

That’s not the case and your TitanXP with compute capability 6.1 is still supported in the latest CUDA 12 release. I’m unsure where you have seen this information about the limited support.

Install the latest release as already mentioned and it will work.

No_Yeah · June 16, 2023, 9:01am

You have saved my times thank you very much sir!