Skip to content

Conversation

NicolasHug
Copy link
Member

@NicolasHug NicolasHug commented Sep 9, 2025

This PR avoids building the C++ video decoder by default and remove the installation of ffmpeg when we build a wheel.
As a result, the "homemade C++" backend of torchvision is removed from official releases. Some of the video decoding interface is still available, but only through an optional dependency on pyav. Eventually we'll want to remove everything, but this PR is an intermediate step that:

  • removes the FFmpeg dependency, which was a major pain point
  • doesn't touch anything on fbcode (which is important, because changes to fbcode are tricky to address).

Copy link

pytorch-bot bot commented Sep 9, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9208

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla bot added the cla signed label Sep 9, 2025
@NicolasHug NicolasHug marked this pull request as ready for review September 9, 2025 12:31
@NicolasHug NicolasHug changed the title Don't build videodecoder by default Don't build video decoder homemade backend by default Sep 9, 2025
@NicolasHug NicolasHug merged commit a8dc530 into pytorch:main Sep 9, 2025
8 of 28 checks passed
@NicolasHug NicolasHug deleted the disable-decoder-build branch September 9, 2025 13:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants