RuntimeError: cuda video backend is not available.

[Caspeerrr]

🐛 Describe the bug

When trying to set the videoreader backend to cuda (torchvision.set_video_backend('cuda')) I get the error below: RuntimeError: cuda video backend is not available. I followed the instructions to use the videoreader on cuda. I.e. I installed pytorch nightly and build torchvision from source. My DockerFile is given below:

FROM nvidia/cuda:11.8.0-devel-ubuntu22.04

RUN apt-get update
RUN apt-get install -y python3-pip
# RUN pip3 install --upgrade pip3

RUN apt-get update
RUN yes | apt install nvidia-cuda-toolkit

RUN pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu118

RUN git clone https://github.com/pytorch/vision.git
WORKDIR "/vision"
RUN python3 setup.py develop

RUN pip3 install ffmpeg-python
RUN pip3 install av --upgrade

As far as I can see the environment has been installed with the expected versions etc. Is this a bug or am I doing something wrong?

Versions

PyTorch version: 2.2.0.dev20231213+cu118 Is debug build: False CUDA used to build PyTorch: 11.8 ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version: Could not collect CMake version: Could not collect Libc version: glibc-2.35

Python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime) Python platform: Linux-5.10.0-25-amd64-x86_64-with-glibc2.35 Is CUDA available: True CUDA runtime version: 11.8.89 CUDA_MODULE_LOADING set to: LAZY Nvidia driver version: 535.146.02 cuDNN version: Could not collect HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True

Versions of relevant libraries: [pip3] numpy==1.24.1 [pip3] pytorch-triton==2.1.0+bcad9dabe1 [pip3] torch==2.2.0.dev20231213+cu118 [pip3] torchaudio==2.2.0.dev20231213+cu118 [pip3] torchvision==0.18.0.dev20231213+cu118 [conda] Could not collect

[NicolasHug]

Hi @Caspeerrr , I'm really sorry but unfortunately, my only recommendation at this point will be to not use the cuda video decoder from torchvision. It's still in BETA stage and there are many edge-cases that we need to figure out before encouraging a wider adoption. At this point time time, I would recommend that you look into the torchaudio video decoder as an alternative

[bhack]

Any update on this? What is the status? Is it still recommended to use torchaudio?