TensorRT icon indicating copy to clipboard operation
TensorRT copied to clipboard

Pip install fails due to deprecated nvidia-cuda-runtime-cu13

Open NevermindNilas opened this issue 2 months ago • 5 comments

Loading personal and system profiles took 722ms.
 python -m venv venv
 cd .\venv\Scripts\
 .\Activate.ps1
 pip install tensorrt
Collecting tensorrt
  Downloading tensorrt-10.13.3.9.tar.gz (40 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Collecting tensorrt_cu13==10.13.3.9 (from tensorrt)
  Downloading tensorrt_cu13-10.13.3.9.tar.gz (18 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
Collecting tensorrt_cu13_libs==10.13.3.9 (from tensorrt_cu13==10.13.3.9->tensorrt)
  Downloading tensorrt_cu13_libs-10.13.3.9.tar.gz (706 bytes)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Collecting tensorrt_cu13_bindings==10.13.3.9 (from tensorrt_cu13==10.13.3.9->tensorrt)
  Downloading tensorrt_cu13_bindings-10.13.3.9-cp313-none-win_amd64.whl.metadata (606 bytes)
Collecting nvidia-cuda-runtime-cu13 (from tensorrt_cu13_libs==10.13.3.9->tensorrt_cu13==10.13.3.9->tensorrt)
  Downloading nvidia_cuda_runtime_cu13-0.0.1.tar.gz (1.4 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Downloading tensorrt_cu13_bindings-10.13.3.9-cp313-none-win_amd64.whl (876 kB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 877.0/877.0 kB 18.5 MB/s  0:00:00
Building wheels for collected packages: tensorrt, tensorrt_cu13, tensorrt_cu13_libs, nvidia-cuda-runtime-cu13
  Building wheel for tensorrt (pyproject.toml) ... done
  Created wheel for tensorrt: filename=tensorrt-10.13.3.9-py2.py3-none-any.whl size=46510 sha256=907e7d959f58d32454429b8b22a55925b3e5472443c96aaa13b95347fcf8d8db
  Stored in directory: c:\users\nilas\appdata\local\pip\cache\wheels\7a\c2\25\ec82345436752fc7d1057ce55bc2e187c601c3fa599fab052d
  Building wheel for tensorrt_cu13 (pyproject.toml) ... done
  Created wheel for tensorrt_cu13: filename=tensorrt_cu13-10.13.3.9-py2.py3-none-any.whl size=17560 sha256=637230bc0359b9430d2225a4e4250c8192d166c17670a2e51eb804f8d51901ba
  Stored in directory: c:\users\nilas\appdata\local\pip\cache\wheels\6d\35\d6\7e3bc31185cdbd140ef7e2a6e4deb8dbca7c8a5a5750a02e26
  Building wheel for tensorrt_cu13_libs (pyproject.toml) ... done
  Created wheel for tensorrt_cu13_libs: filename=tensorrt_cu13_libs-10.13.3.9-py2.py3-none-win_amd64.whl size=1327112482 sha256=2561ed2f0a79c011701c9e74e0c5ccac5f84c19f5c51d17986b55d9adb4f2d21
  Stored in directory: c:\users\nilas\appdata\local\pip\cache\wheels\8a\50\31\96014fb119e9efdcd12c5e48b810582cd7e362945455ed4ae8
  Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml) ... error
  error: subprocess-exited-with-error

  × Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [10 lines of output]
      \u26a0\ufe0f THIS PROJECT 'nvidia-cuda-runtime-cu13' IS DEPRECATED.
      Please use 'nvidia-cuda-runtime' instead.

      To install the correct package, use:

          pip install nvidia-cuda-runtime

      For more information, visit: https://pypi.org/project/nvidia-cuda-runtime/


      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for nvidia-cuda-runtime-cu13
Successfully built tensorrt tensorrt_cu13 tensorrt_cu13_libs
Failed to build nvidia-cuda-runtime-cu13

[notice] A new release of pip is available: 25.2 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
error: failed-wheel-build-for-install

× Failed to build installable wheels for some pyproject.toml based projects
╰─> nvidia-cuda-runtime-cu13

This is in a blank venv and just simply running pip install tensorrt as seen in the logs.

Trying to work around it with a pip install nvidia-cuda-runtime and then trying to pip install tensorrt doesn't fix the issue either, probably due to the hardcoded link to the -cu13 library.

 pip install nvidia-cuda-runtime
Collecting nvidia-cuda-runtime
  Downloading nvidia_cuda_runtime-13.0.96-py3-none-win_amd64.whl.metadata (1.7 kB)
Downloading nvidia_cuda_runtime-13.0.96-py3-none-win_amd64.whl (2.9 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.9/2.9 MB 39.9 MB/s  0:00:00
Installing collected packages: nvidia-cuda-runtime
Successfully installed nvidia-cuda-runtime-13.0.96

[notice] A new release of pip is available: 25.2 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
 pip install tensorrt
Collecting tensorrt
  Using cached tensorrt-10.13.3.9-py2.py3-none-any.whl
Collecting tensorrt_cu13==10.13.3.9 (from tensorrt)
  Using cached tensorrt_cu13-10.13.3.9-py2.py3-none-any.whl
Collecting tensorrt_cu13_libs==10.13.3.9 (from tensorrt_cu13==10.13.3.9->tensorrt)
  Using cached tensorrt_cu13_libs-10.13.3.9-py2.py3-none-win_amd64.whl
Collecting tensorrt_cu13_bindings==10.13.3.9 (from tensorrt_cu13==10.13.3.9->tensorrt)
  Using cached tensorrt_cu13_bindings-10.13.3.9-cp313-none-win_amd64.whl.metadata (606 bytes)
Collecting nvidia-cuda-runtime-cu13 (from tensorrt_cu13_libs==10.13.3.9->tensorrt_cu13==10.13.3.9->tensorrt)
  Using cached nvidia_cuda_runtime_cu13-0.0.1.tar.gz (1.4 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Using cached tensorrt_cu13_bindings-10.13.3.9-cp313-none-win_amd64.whl (876 kB)
Building wheels for collected packages: nvidia-cuda-runtime-cu13
  Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml) ... error
  error: subprocess-exited-with-error

  × Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [10 lines of output]
      \u26a0\ufe0f THIS PROJECT 'nvidia-cuda-runtime-cu13' IS DEPRECATED.
      Please use 'nvidia-cuda-runtime' instead.

      To install the correct package, use:

          pip install nvidia-cuda-runtime

      For more information, visit: https://pypi.org/project/nvidia-cuda-runtime/


      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for nvidia-cuda-runtime-cu13
Failed to build nvidia-cuda-runtime-cu13

[notice] A new release of pip is available: 25.2 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
error: failed-wheel-build-for-install

× Failed to build installable wheels for some pyproject.toml based projects
╰─> nvidia-cuda-runtime-cu13

Steps to reproduce: pip install tensorrt

System specs: 13700k 3090 latest drivers through the NVidia APP. Windows 11 24h2

NevermindNilas avatar Nov 02 '25 09:11 NevermindNilas

for added context running pip install tensorrt==10.13.2.6 fixes the issue as it doesn't seem to rely on nvidia-cuda-runtime-cu13

 pip install tensorrt==10.13.2.6
Collecting tensorrt==10.13.2.6
  Downloading tensorrt-10.13.2.6.tar.gz (40 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Requirement already satisfied: tensorrt_cu12==10.13.2.6 in f:\test\venv\lib\site-packages (from tensorrt==10.13.2.6) (10.13.2.6)
Requirement already satisfied: tensorrt_cu12_libs==10.13.2.6 in f:\test\venv\lib\site-packages (from tensorrt_cu12==10.13.2.6->tensorrt==10.13.2.6) (10.13.2.6)
Requirement already satisfied: tensorrt_cu12_bindings==10.13.2.6 in f:\test\venv\lib\site-packages (from tensorrt_cu12==10.13.2.6->tensorrt==10.13.2.6) (10.13.2.6)
Requirement already satisfied: nvidia-cuda-runtime-cu12 in f:\test\venv\lib\site-packages (from tensorrt_cu12_libs==10.13.2.6->tensorrt_cu12==10.13.2.6->tensorrt==10.13.2.6) (12.9.79)
Building wheels for collected packages: tensorrt
  Building wheel for tensorrt (pyproject.toml) ... done
  Created wheel for tensorrt: filename=tensorrt-10.13.2.6-py2.py3-none-any.whl size=46543 sha256=fef54b1d453e093ba064cf377088b81a365c5253aae95f4162c661c616f8a058
  Stored in directory: c:\users\nilas\appdata\local\pip\cache\wheels\49\5e\56\4b9680e2ed0c05c2c3e43e12e5c8598edf70d106c10ab677c3
Successfully built tensorrt
Installing collected packages: tensorrt
Successfully installed tensorrt-10.13.2.6

[notice] A new release of pip is available: 25.2 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip

So it is specific to the latest wheel, tensorrt==10.13.3.9

NevermindNilas avatar Nov 02 '25 09:11 NevermindNilas

In your case, 10.13.2.6 works because it's using cu12. If I force 10.13.2.6 to use cu13, the problem is the same. So I am not able to use either version with cu13...

jmspereira avatar Nov 03 '25 09:11 jmspereira

i get same error for tensorrt==10.13.2.6:

#8 1.389 Collecting tensorrt_cu13==10.13.2.6 (from tensorrt==10.13.2.6->-r /tmp/requirements.txt (line 5))
#8 1.400   Downloading tensorrt_cu13-10.13.2.6.tar.gz (18 kB)
#8 1.406   Preparing metadata (setup.py): started
#8 1.585   Preparing metadata (setup.py): finished with status 'done'
#8 1.609 Collecting tensorrt_cu13_bindings==10.13.2.6 (from tensorrt_cu13==10.13.2.6->tensorrt==10.13.2.6->-r /tmp/requirements.txt (line 5))
#8 1.619   Downloading tensorrt_cu13_bindings-10.13.2.6-cp312-none-manylinux_2_28_x86_64.whl.metadata (606 bytes)
#8 1.641 Collecting tensorrt_cu13_libs==10.13.2.6 (from tensorrt_cu13==10.13.2.6->tensorrt==10.13.2.6->-r /tmp/requirements.txt (line 5))
#8 1.652   Downloading tensorrt_cu13_libs-10.13.2.6.tar.gz (704 bytes)
#8 1.656   Installing build dependencies: started
#8 2.437   Installing build dependencies: finished with status 'done'
#8 2.438   Getting requirements to build wheel: started
#8 2.530   Getting requirements to build wheel: finished with status 'done'
#8 2.531   Preparing metadata (pyproject.toml): started
#8 18.68   Preparing metadata (pyproject.toml): finished with status 'done'
#8 18.70 Collecting nvidia-cuda-runtime-cu13 (from tensorrt_cu13_libs==10.13.2.6->tensorrt_cu13==10.13.2.6->tensorrt==10.13.2.6->-r /tmp/requirements.txt (line 5))
#8 18.72   Downloading nvidia_cuda_runtime_cu13-0.0.1.tar.gz (1.4 kB)
#8 18.72   Installing build dependencies: started
#8 20.39   Installing build dependencies: finished with status 'done'
#8 20.39   Getting requirements to build wheel: started
#8 20.55   Getting requirements to build wheel: finished with status 'done'
#8 20.55   Preparing metadata (pyproject.toml): started
#8 20.71   Preparing metadata (pyproject.toml): finished with status 'done'
#8 20.77 Collecting humanfriendly>=9.1 (from coloredlogs->onnxruntime-gpu==1.22.0->-r /tmp/requirements.txt (line 1))
#8 20.78   Downloading humanfriendly-10.0-py2.py3-none-any.whl.metadata (9.2 kB)
#8 20.81 Collecting mpmath<1.4,>=1.1.0 (from sympy->onnxruntime-gpu==1.22.0->-r /tmp/requirements.txt (line 1))
#8 20.82   Downloading mpmath-1.3.0-py3-none-any.whl.metadata (8.6 kB)
#8 20.84 Downloading onnxruntime_gpu-1.22.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (283.2 MB)
#8 22.90    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 283.2/283.2 MB 23.6 MB/s eta 0:00:00
#8 22.91 Downloading onnx-1.18.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (17.6 MB)
#8 23.01    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 17.6/17.6 MB 151.8 MB/s eta 0:00:00
#8 23.02 Downloading colored-2.3.0-py3-none-any.whl (18 kB)
#8 23.04 Downloading polygraphy-0.49.24-py2.py3-none-any.whl (362 kB)
#8 23.04    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 362.4/362.4 kB 106.5 MB/s eta 0:00:00
#8 23.13 Downloading tensorrt_cu13_bindings-10.13.2.6-cp312-none-manylinux_2_28_x86_64.whl (1.2 MB)
#8 23.14    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 148.8 MB/s eta 0:00:00
#8 23.15 Downloading protobuf-6.33.0-cp39-abi3-manylinux2014_x86_64.whl (323 kB)
#8 23.15    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 323.2/323.2 kB 97.7 MB/s eta 0:00:00
#8 23.16 Downloading coloredlogs-15.0.1-py2.py3-none-any.whl (46 kB)
#8 23.17    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 46.0/46.0 kB 22.9 MB/s eta 0:00:00
#8 23.18 Downloading flatbuffers-25.9.23-py2.py3-none-any.whl (30 kB)
#8 23.19 Downloading sympy-1.14.0-py3-none-any.whl (6.3 MB)
#8 23.23    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.3/6.3 MB 177.8 MB/s eta 0:00:00
#8 23.24 Downloading humanfriendly-10.0-py2.py3-none-any.whl (86 kB)
#8 23.24    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 86.8/86.8 kB 41.2 MB/s eta 0:00:00
#8 23.25 Downloading mpmath-1.3.0-py3-none-any.whl (536 kB)
#8 23.26    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 536.2/536.2 kB 119.0 MB/s eta 0:00:00
#8 23.60 Building wheels for collected packages: tensorrt, tensorrt_cu13, tensorrt_cu13_libs, nvidia-cuda-runtime-cu13
#8 23.60   Building wheel for tensorrt (setup.py): started
#8 23.84   Building wheel for tensorrt (setup.py): finished with status 'done'
#8 23.84   Created wheel for tensorrt: filename=tensorrt-10.13.2.6-py2.py3-none-any.whl size=46433 sha256=da2fb7d4608612dfd5086f5f15fca72edfb2d3acb767b4b7ec1adc668fc568df
#8 23.84   Stored in directory: /root/.cache/pip/wheels/5a/22/e7/4d28e4907a118d0a014adfec88205918792bb11cd061646091
#8 23.84   Building wheel for tensorrt_cu13 (setup.py): started
#8 24.08   Building wheel for tensorrt_cu13 (setup.py): finished with status 'done'
#8 24.08   Created wheel for tensorrt_cu13: filename=tensorrt_cu13-10.13.2.6-py2.py3-none-any.whl size=17432 sha256=bc143f63eb8bf42c9a3464cce7c943a9e3271a989a450647ab3419bd1530a951
#8 24.08   Stored in directory: /root/.cache/pip/wheels/de/a4/19/4b06ee6d941601c0795594f260be07c42532c8a4a24cf6a1ff
#8 24.08   Building wheel for tensorrt_cu13_libs (pyproject.toml): started
#8 25.91   Building wheel for tensorrt_cu13_libs (pyproject.toml): finished with status 'done'
#8 27.67   Created wheel for tensorrt_cu13_libs: filename=tensorrt_cu13_libs-10.13.2.6-py2.py3-none-manylinux_2_28_x86_64.whl size=2740427176 sha256=044468b0705728c0c8f0e95325ba881ff3e23bd2eebce0b1e628953ff90bbb5f
#8 27.67   Stored in directory: /root/.cache/pip/wheels/fe/38/12/357ce01af52af73540a307d0f500629f7683148385297f336b
#8 27.67   Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml): started
#8 27.79   Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml): finished with status 'error'
#8 27.79   error: subprocess-exited-with-error
#8 27.79   
#8 27.79   × Building wheel for nvidia-cuda-runtime-cu13 (pyproject.toml) did not run successfully.
#8 27.79   │ exit code: 1
#8 27.79   ╰─> [10 lines of output]
#8 27.79       ⚠️ THIS PROJECT 'nvidia-cuda-runtime-cu13' IS DEPRECATED.

tensorrt==10.13.2.6 └─ tensorrt_cu13==10.13.2.6 ├─ tensorrt_cu13_libs==10.13.2.6 │ └─ nvidia-cuda-runtime-cu13 ← deprecated / fails to build └─ tensorrt_cu13_bindings==10.13.2.6

cc @lix19937 @MrGeva @kevinch-nv

geraldstanje1 avatar Nov 03 '25 22:11 geraldstanje1

Still not working

Alan-LanFeng avatar Nov 07 '25 10:11 Alan-LanFeng

Any updates?... version 10.13.2.6 still cannot be installed

bobhash avatar Nov 20 '25 09:11 bobhash