update python and ffmpeg version in docker cuda #5582
Conversation
dongfengweixiao
force-pushed
the
update_python_and_ffmpeg_in_docker_cuda
branch
from
f57d796 to
a07900c
Compare
|
Does this reduce the need for transcoding in some cases? |
1. update python version to 3.12(the version in the official repository is 3.10, which encountered errors while executing some tasks) 2. update ffmpeg version to 6.0(the version in the official repository is 4. x, which does not meet the running requirements of stash. The downloaded ffmpeg does not support Nvidia's CUDA.) 3. preinstall cloudscraper bs4 stashapp tools lxml using pip Signed-off-by: Dee.H.Y <[email protected]>
dongfengweixiao
force-pushed
the
update_python_and_ffmpeg_in_docker_cuda
branch
from
a07900c to
09d8be6
Compare
|
Python3.13 should be used, and you don't address the CUDA upgrade arg you are also statically linking ffmpeg when the one that is downloaded by stash already supports nvenc/ CUDA https://johnvansickle.com/ffmpeg/ |
It appears unnecessary to upgrade to Python 3.13 since all plugins function well with Python 3.12. Could you provide any links that specifically require Python 3.13?
I apologize for not grasping your intention. Could you please provide a more detailed explanation?
I’m sorry to report that despite trying the version you suggested, it doesn’t seem to run smoothly within the Docker environment. The compatibility of FFmpeg in Docker is dependent on the NVIDIA driver version within the container, which in turn is linked to the NVIDIA driver version on the host machine. After testing various versions of FFmpeg, I’ve confirmed that this particular version functions as expected. |
Other suggestions:
Adding the cuda version to cuda build suffix in the makefile is crucial.
cuda requires that the version in the container is not higher than the version in the host computer.