Github action to publish a CUDA build alongside the alpine one

[NodudeWasTaken]

For some time people complained that there was no official cuda build.
So i changed the build a little bit and now there could be.

Comments? Do we want to do this? Did i use too many newlines? Is anyone actually gonna use the NVidia arm64 build?
There are also some people who have a jellyfin-ffmpeg build which could be interesting, but for now i just made it similar to the old one.

[kermieisinthehouse]

Have you tested that the nvenc restriction is actually lifted? I would think that this would need to be done in the host's Nvidia driver, not from inside the container, but I could be wrong.

[NodudeWasTaken]

Have you tested that the nvenc restriction is actually lifted? I would think that this would need to be done in the host's Nvidia driver, not from inside the container, but I could be wrong.

You know, i actually didnt, i dont usually watch more than 3 porno's at the same time 😋
But the author claims it does, and i can confirm that it also works in my docker.

ps aux confirms its using h264_nvenc, so it works fine.

[feederbox826]

bit of a necro, but

• the current dockerfile encapsulates both non-CUDA/non-hwaccel and the cuda/hwaccel versions, this greatly inflates all image sizes
• I'm hoping to address this once and for all in [Feature] Docker container overhaul #4300 with flavored builds for jellyfin-ffmpeg and pre-compiled driverless ffmpeg builds that can still use hwaccel

as for the nvidia patch, one of the explicit goals is to patch on hosts without patched drivers (https://github.com/keylase/nvidia-patch/tree/master#docker-support)

[NodudeWasTaken]

bit of a necro, but

* the current dockerfile encapsulates both non-CUDA/non-hwaccel and the cuda/hwaccel versions, this greatly inflates all image sizes

* I'm hoping to address this once and for all in [[Feature] Docker container overhaul #4300](https://github.com/stashapp/stash/issues/4300) with flavored builds for jellyfin-ffmpeg and pre-compiled driverless ffmpeg builds that can still use hwaccel

as for the nvidia patch, one of the explicit goals is to patch on hosts without patched drivers (https://github.com/keylase/nvidia-patch/tree/master#docker-support)

Guess i'll close in favor of yours