RuntimeError in Step3 of Clip Extracting

[Wang-Shengyuan]

Thank you for your work.

The dataset is useful. However, when I tried the Step3 of Clip Extracting with the command of python 3_nvtranscoding.py --input_clip_dir sekai-real-walking-hq --input_video_dir ./videos --output_dir ./vstreams,

these appears an error of RuntimeError: NvEncoder : m_nvenc.nvEncOpenEncodeSessionEx(&encodeSessionExParams, &hEncoder) returned error 2 at /tmp/pip-req-build-7a0a3q6t/src/VideoCodecSDKUtils/helper_classes/NvCodec/NvEncoder/NvEncoder.cpp:69

Could you look into it? This seems to be an error from the script. Thank you so much.

[Lixsp11]

If you are running inside an NVIDIA Docker container, note that the NVIDIA libraries are mounted selectively according to the NVIDIA_DRIVER_CAPABILITIES setting, see https://github.com/NVIDIA/nvidia-container-runtime#nvidia_driver_capabilities and https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/user-guide.html#environment-variables-oci-spec.

To make the encoding libraries available, simply start the container with the environment variable NVIDIA_DRIVER_CAPABILITIES=all.

[Wang-Shengyuan]

Hi there,

Thank you for your quick response. Unfortunately, I’m not using an NVIDIA Docker container, so that solution doesn’t work in my setup.

I’ve implemented an alternative approach with FFmpeg in place of the original NVIDIA encoder and opened a pull request. Could you please confirm whether this is functionally equivalent to the official implementation?

Thanks!

[Lixsp11]

Closed due to no response in #7. Feel free to reopen if you have more info.