I ran the following conda command for installing pytorch - conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
The output I got is: Collecting package metadata (current_repodata.json): done
==> WARNING: A newer version of conda exists. <==
Please update conda by running
$ conda update -n base -c defaults conda
environment location: C:\Users\ranga\anaconda3\envs\newml
added / updated specs:
The following packages will be downloaded:
package | build
---------------------------|-----------------
cuda-cccl-12.1.55 | 0 1.2 MB nvidia
cuda-cudart-11.8.89 | 0 1.4 MB nvidia
cuda-cudart-dev-11.8.89 | 0 723 KB nvidia
cuda-cupti-11.8.87 | 0 11.5 MB nvidia
cuda-libraries-11.8.0 | 0 1 KB nvidia
cuda-libraries-dev-11.8.0 | 0 1 KB nvidia
cuda-nvrtc-11.8.89 | 0 72.1 MB nvidia
cuda-nvrtc-dev-11.8.89 | 0 16.1 MB nvidia
cuda-nvtx-11.8.86 | 0 43 KB nvidia
cuda-profiler-api-12.1.55 | 0 18 KB nvidia
libcublas-11.11.3.6 | 0 33 KB nvidia
libcublas-dev-11.11.3.6 | 0 375.9 MB nvidia
libcufft-10.9.0.58 | 0 6 KB nvidia
libcufft-dev-10.9.0.58 | 0 144.6 MB nvidia
libcurand-10.3.2.56 | 0 3 KB nvidia
libcurand-dev-10.3.2.56 | 0 50.0 MB nvidia
libcusolver-11.4.1.48 | 0 29 KB nvidia
libcusolver-dev-11.4.1.48 | 0 94.1 MB nvidia
libcusparse-11.7.5.86 | 0 13 KB nvidia
libcusparse-dev-11.7.5.86 | 0 175.7 MB nvidia
libnpp-11.8.0.86 | 0 294 KB nvidia
libnpp-dev-11.8.0.86 | 0 143.2 MB nvidia
libnvjpeg-11.9.0.86 | 0 4 KB nvidia
libnvjpeg-dev-11.9.0.86 | 0 1.9 MB nvidia
------------------------------------------------------------
Total: 1.06 GB
The following NEW packages will be INSTALLED:
blas pkgs/main/win-64::blas-1.0-mkl
Proceed ([y]/n)? y
Downloading and Extracting Packages
I am still getting the same error “Torch not compiled with CUDA enabled.”
I cannot reproduce the issue on a Windows system using the same command so I guess your environment might have multiple PyTorch binaries installed (and the CPU-only one is picked),
conda install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia
Collecting package metadata (current_repodata.json): done
...
added / updated specs:
- pytorch
- pytorch-cuda=11.8
- torchvision
The following packages will be downloaded:
package | build
---------------------------|-----------------
brotlipy-0.7.0 |py38h2bbff1b_1003 412 KB
cffi-1.15.1 | py38h2bbff1b_3 238 KB
cryptography-39.0.1 | py38h21b164f_0 1.0 MB
cuda-cccl-12.1.55 | 0 1.2 MB nvidia
cuda-cudart-dev-11.8.89 | 0 723 KB nvidia
cuda-cupti-11.8.87 | 0 11.5 MB nvidia
cuda-libraries-dev-11.8.0 | 0 1 KB nvidia
cuda-nvrtc-dev-11.8.89 | 0 16.1 MB nvidia
cuda-profiler-api-12.1.55 | 0 18 KB nvidia
cuda-runtime-11.8.0 | 0 1 KB nvidia
filelock-3.9.0 | py38haa95532_0 19 KB
flit-core-3.8.0 | py38haa95532_0 85 KB
freetype-2.12.1 | ha860e81_0 490 KB
giflib-5.2.1 | h8cc25b3_3 88 KB
idna-3.4 | py38haa95532_0 94 KB
jinja2-3.1.2 | py38haa95532_0 211 KB
jpeg-9e | h2bbff1b_1 320 KB
libcublas-dev-11.11.3.6 | 0 375.9 MB nvidia
libcufft-dev-10.9.0.58 | 0 144.6 MB nvidia
libcurand-10.3.2.56 | 0 3 KB nvidia
libcurand-dev-10.3.2.56 | 0 50.0 MB nvidia
libcusolver-dev-11.4.1.48 | 0 94.1 MB nvidia
libcusparse-dev-11.7.5.86 | 0 175.7 MB nvidia
libdeflate-1.17 | h2bbff1b_0 151 KB
libnpp-dev-11.8.0.86 | 0 143.2 MB nvidia
libpng-1.6.39 | h8cc25b3_0 369 KB
libtiff-4.5.0 | h6c2663c_2 1.2 MB
libuv-1.44.2 | h2bbff1b_0 288 KB
libwebp-1.2.4 | hbc33d0d_1 73 KB
libwebp-base-1.2.4 | h2bbff1b_1 304 KB
lz4-c-1.9.4 | h2bbff1b_0 143 KB
markupsafe-2.1.1 | py38h2bbff1b_0 26 KB
mkl-service-2.4.0 | py38h2bbff1b_0 51 KB
mkl_fft-1.3.1 | py38h277e83a_0 139 KB
mkl_random-1.2.2 | py38hf11a4ad_0 225 KB
networkx-2.8.4 | py38haa95532_1 2.6 MB
numpy-1.23.5 | py38h3b20f71_0 11 KB
numpy-base-1.23.5 | py38h4da318b_0 6.0 MB
pillow-9.4.0 | py38hd77b12b_0 1015 KB
pyopenssl-23.0.0 | py38haa95532_0 97 KB
pysocks-1.7.1 | py38haa95532_0 31 KB
pytorch-2.0.0 |py3.8_cuda11.8_cudnn8_0 1.35 GB pytorch
pytorch-cuda-11.8 | h24eeafa_3 7 KB pytorch
pytorch-mutex-1.0 | cuda 3 KB pytorch
requests-2.28.1 | py38haa95532_1 98 KB
torchvision-0.15.0 | py38_cu118 7.7 MB pytorch
urllib3-1.26.15 | py38haa95532_0 194 KB
win_inet_pton-1.1.0 | py38haa95532_0 35 KB
xz-5.2.10 | h8cc25b3_1 520 KB
zlib-1.2.13 | h8cc25b3_0 113 KB
zstd-1.5.4 | hd43e919_0 683 KB
------------------------------------------------------------
Total: 2.37 GB
The following NEW packages will be INSTALLED:
blas pkgs/main/win-64::blas-1.0-mkl None
brotlipy pkgs/main/win-64::brotlipy-0.7.0-py38h2bbff1b_1003 None
cffi pkgs/main/win-64::cffi-1.15.1-py38h2bbff1b_3 None
charset-normalizer pkgs/main/noarch::charset-normalizer-2.0.4-pyhd3eb1b0_0 None
cryptography pkgs/main/win-64::cryptography-39.0.1-py38h21b164f_0 None
cuda-cccl nvidia/win-64::cuda-cccl-12.1.55-0 None
cuda-cudart nvidia/win-64::cuda-cudart-11.8.89-0 None
cuda-cudart-dev nvidia/win-64::cuda-cudart-dev-11.8.89-0 None
cuda-cupti nvidia/win-64::cuda-cupti-11.8.87-0 None
cuda-libraries nvidia/win-64::cuda-libraries-11.8.0-0 None
cuda-libraries-dev nvidia/win-64::cuda-libraries-dev-11.8.0-0 None
cuda-nvrtc nvidia/win-64::cuda-nvrtc-11.8.89-0 None
cuda-nvrtc-dev nvidia/win-64::cuda-nvrtc-dev-11.8.89-0 None
cuda-nvtx nvidia/win-64::cuda-nvtx-11.8.86-0 None
cuda-profiler-api nvidia/win-64::cuda-profiler-api-12.1.55-0 None
cuda-runtime nvidia/win-64::cuda-runtime-11.8.0-0 None
filelock pkgs/main/win-64::filelock-3.9.0-py38haa95532_0 None
flit-core pkgs/main/win-64::flit-core-3.8.0-py38haa95532_0 None
freetype pkgs/main/win-64::freetype-2.12.1-ha860e81_0 None
giflib pkgs/main/win-64::giflib-5.2.1-h8cc25b3_3 None
idna pkgs/main/win-64::idna-3.4-py38haa95532_0 None
intel-openmp pkgs/main/win-64::intel-openmp-2021.4.0-haa95532_3556 None
jinja2 pkgs/main/win-64::jinja2-3.1.2-py38haa95532_0 None
jpeg pkgs/main/win-64::jpeg-9e-h2bbff1b_1 None
lerc pkgs/main/win-64::lerc-3.0-hd77b12b_0 None
libcublas nvidia/win-64::libcublas-11.11.3.6-0 None
libcublas-dev nvidia/win-64::libcublas-dev-11.11.3.6-0 None
libcufft nvidia/win-64::libcufft-10.9.0.58-0 None
libcufft-dev nvidia/win-64::libcufft-dev-10.9.0.58-0 None
libcurand nvidia/win-64::libcurand-10.3.2.56-0 None
libcurand-dev nvidia/win-64::libcurand-dev-10.3.2.56-0 None
libcusolver nvidia/win-64::libcusolver-11.4.1.48-0 None
libcusolver-dev nvidia/win-64::libcusolver-dev-11.4.1.48-0 None
libcusparse nvidia/win-64::libcusparse-11.7.5.86-0 None
libcusparse-dev nvidia/win-64::libcusparse-dev-11.7.5.86-0 None
libdeflate pkgs/main/win-64::libdeflate-1.17-h2bbff1b_0 None
libnpp nvidia/win-64::libnpp-11.8.0.86-0 None
libnpp-dev nvidia/win-64::libnpp-dev-11.8.0.86-0 None
libnvjpeg nvidia/win-64::libnvjpeg-11.9.0.86-0 None
libnvjpeg-dev nvidia/win-64::libnvjpeg-dev-11.9.0.86-0 None
libpng pkgs/main/win-64::libpng-1.6.39-h8cc25b3_0 None
libtiff pkgs/main/win-64::libtiff-4.5.0-h6c2663c_2 None
libuv pkgs/main/win-64::libuv-1.44.2-h2bbff1b_0 None
libwebp pkgs/main/win-64::libwebp-1.2.4-hbc33d0d_1 None
libwebp-base pkgs/main/win-64::libwebp-base-1.2.4-h2bbff1b_1 None
lz4-c pkgs/main/win-64::lz4-c-1.9.4-h2bbff1b_0 None
markupsafe pkgs/main/win-64::markupsafe-2.1.1-py38h2bbff1b_0 None
mkl pkgs/main/win-64::mkl-2021.4.0-haa95532_640 None
mkl-service pkgs/main/win-64::mkl-service-2.4.0-py38h2bbff1b_0 None
mkl_fft pkgs/main/win-64::mkl_fft-1.3.1-py38h277e83a_0 None
mkl_random pkgs/main/win-64::mkl_random-1.2.2-py38hf11a4ad_0 None
mpmath pkgs/main/win-64::mpmath-1.2.1-py38haa95532_0 None
networkx pkgs/main/win-64::networkx-2.8.4-py38haa95532_1 None
numpy pkgs/main/win-64::numpy-1.23.5-py38h3b20f71_0 None
numpy-base pkgs/main/win-64::numpy-base-1.23.5-py38h4da318b_0 None
pillow pkgs/main/win-64::pillow-9.4.0-py38hd77b12b_0 None
pycparser pkgs/main/noarch::pycparser-2.21-pyhd3eb1b0_0 None
pyopenssl pkgs/main/win-64::pyopenssl-23.0.0-py38haa95532_0 None
pysocks pkgs/main/win-64::pysocks-1.7.1-py38haa95532_0 None
pytorch pytorch/win-64::pytorch-2.0.0-py3.8_cuda11.8_cudnn8_0 None
pytorch-cuda pytorch/win-64::pytorch-cuda-11.8-h24eeafa_3 None
pytorch-mutex pytorch/noarch::pytorch-mutex-1.0-cuda None
requests pkgs/main/win-64::requests-2.28.1-py38haa95532_1 None
six pkgs/main/noarch::six-1.16.0-pyhd3eb1b0_1 None
sympy pkgs/main/win-64::sympy-1.11.1-py38haa95532_0 None
tk pkgs/main/win-64::tk-8.6.12-h2bbff1b_0 None
torchvision pytorch/win-64::torchvision-0.15.0-py38_cu118 None
typing_extensions pkgs/main/win-64::typing_extensions-4.4.0-py38haa95532_0 None
urllib3 pkgs/main/win-64::urllib3-1.26.15-py38haa95532_0 None
win_inet_pton pkgs/main/win-64::win_inet_pton-1.1.0-py38haa95532_0 None
xz pkgs/main/win-64::xz-5.2.10-h8cc25b3_1 None
zlib pkgs/main/win-64::zlib-1.2.13-h8cc25b3_0 None
zstd pkgs/main/win-64::zstd-1.5.4-hd43e919_0 None
...
>>> import torch
>>> torch.version.cuda
'11.8'
>>> torch.__version__
'2.0.0'
>>> x = torch.randn(1).cuda()
>>> x
tensor([1.7900], device='cuda:0')
I did all of this in separate conda environments; so there is no chance of two versions of PyTorch binaries existing. According to this blog - How to setup PyTorch with CUDA in Windows 11 | by Nilotpal Sinha PhD | Medium ; we need to first install the CUDA toolkit version based on the CUDA version specified in the output of nvidia-smi.
Do you think since the RTX 4060 has CUDA 12 specified in it (based on nvidia-smi) and PyTorch does not yet provide support for this version is causing this issue. Is there any way I can force RTX 4060 to make use of a specific CUDA version - say 11.8.
That’s not the case, since the binaries use their own CUDA runtime and libs as already mentioned.
No, since GPUs with compute capability 8.9 are supported in the current releases.
Yes, you would have to install the corresponding binaries as shown in my example.
With that being said, could you execute the three lines of code from my previous code snippet and post the result?
This is the output I got for running the command specified:
conda install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia
Collecting package metadata (current_repodata.json): done
==> WARNING: A newer version of conda exists. <==
Please update conda by running
$ conda update -n base -c defaults conda
environment location: C:\Users\ranga\anaconda3\envs\newml3
added / updated specs:
The following NEW packages will be INSTALLED:
blas pkgs/main/win-64::blas-1.0-mkl
Proceed ([y]/n)? y
Preparing transaction: done
Hi,
Actually, the issue is fixed now, I tried to switch to venv environment as somone suggested above and it simply worked. Perhaps conda is not pulling the version it is supposed to pull although it seems to label the pytorch correctly
jvnagao
June 13, 2023, 8:59am
28
if you tried everything and can’t get it, inside terminal of Pycharm IDE or Jupyter Notebook cell, run this code:
Pycharmhttps://download.pytorch.org/whl/cu118
Jupyter Notebookhttps://download.pytorch.org/whl/cu118
test your code again
