jrottenberg
CUDA_ERROR_COMPAT_NOT_SUPPORTED_ON_DEVICE
Hi Team,
thx for this great effort. We are using following cmd to try encoding with nvidia hw support.
docker run --gpus all --runtime=nvidia --rm -v $(pwd):$(pwd) -w $(pwd) jrottenberg/ffmpeg:7.0.1-nvidia2204 -i $(pwd)/t.mkv -c:v h264_nvenc -r 23.976 -g 48 -keyint_min 48 -crf 26 -c:a aac -ac 2 -ar 44100 -b:a 64k -bsf:v h264_mp4toannexb -f hls -hls_time 3 -hls_segment_size 4000000 -hls_playlist_type vod -hls_flags independent_segments $(pwd)/t.m3u8
and it always got terminated with following error:
[h264_nvenc @ 0x563d07d84580] dl_fn->cuda_dl->cuInit(0) failed -> CUDA_ERROR_COMPAT_NOT_SUPPORTED_ON_DEVICE: forward compatibility was attempted on non supported HW [vost#0:0/h264_nvenc @ 0x563d07d81600] Error while opening encoder - maybe incorrect parameters such as bit_rate, rate, width or height. [vf#0:0 @ 0x563d07d855c0] Error sending frames to consumers: Unknown error occurred [vf#0:0 @ 0x563d07d855c0] Task finished with error code: -1313558101 (Unknown error occurred) [vf#0:0 @ 0x563d07d855c0] Terminating thread with return code -1313558101 (Unknown error occurred) [vost#0:0/h264_nvenc @ 0x563d07d81600] Could not open encoder before EOF [vost#0:0/h264_nvenc @ 0x563d07d81600] Task finished with error code: -22 (Invalid argument) [vost#0:0/h264_nvenc @ 0x563d07d81600] Terminating thread with return code -22 (Invalid argument) [out#0/hls @ 0x563d07dd90c0] Nothing was written into output file, because at least one of its streams received no packets. frame= 0 fps=0.0 q=0.0 Lsize= 0KiB time=N/A bitrate=N/A speed=N/A [aac @ 0x563d08c6f540] Qavg: 183.388 Conversion failed!
the system got alreadyt nvidia-docker and driver installed
any ideas why?
This can also be due to differing versions of the Cuda between host and container.
I had the same issue and it was due to using an older (12.2.1) version of Cuda provided by my distro on my host system, while these images use a newer version of the nVidia image.
A possible fix is to clone this repo, modify the Dockerfile to base on the same major/minor nvidia:cuda -devel and -runtime images, and build your stack from there. I did this and it solved my issue. Upgrading your host's Cuda stack to match at least the version used in the images may also work.
