Skip to content

[wip] Lossless hdr#1311

Draft
mollyxu wants to merge 22 commits intometa-pytorch:mainfrom
mollyxu:lossless-hdr
Draft

[wip] Lossless hdr#1311
mollyxu wants to merge 22 commits intometa-pytorch:mainfrom
mollyxu:lossless-hdr

Conversation

@mollyxu
Copy link
Copy Markdown
Contributor

@mollyxu mollyxu commented Mar 25, 2026

No description provided.

@pytorch-bot
Copy link
Copy Markdown

pytorch-bot bot commented Mar 25, 2026

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/meta-pytorch/torchcodec/1311

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

❌ 2 New Failures

As of commit 230896a with merge base bdccc0b (image):

NEW FAILURES - The following jobs have failed:

  • Build and test Linux wheel / install-and-test-third-party-interface (3.10, 8.0) (gh)
    /home/runner/work/torchcodec/torchcodec/test/third-party-interface/ThirdPartyInterfaceTest.cpp:37:5: error: could not convert ‘<lambda closure object>facebook::torchcodec::{anonymous}::<lambda(const facebook::torchcodec::StableDevice&)>()’ from ‘facebook::torchcodec::{anonymous}::<lambda(const facebook::torchcodec::StableDevice&)>’ to ‘facebook::torchcodec::CreateDeviceInterfaceFn’ {aka ‘std::function<facebook::torchcodec::DeviceInterface*(const torch::stable::Device&)>’}
  • Build and test Windows wheel / install-and-test (3.10, 4.4.2) (gh)
    test/test_decoders.py::TestVideoDecoder::test_10bit_float32_against_ffmpeg_rgb48[approximate-asset2-cpu]

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

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Mar 25, 2026
@mollyxu mollyxu changed the title Lossless hdr [wip] Lossless hdr Mar 25, 2026
@s-deeper
Copy link
Copy Markdown

s-deeper commented Apr 1, 2026

Small correctness concern in the new P016 path:

convertP016FrameToRGB() always falls back to bt601LimitedRange16ColorTwist for non-BT.709 / non-BT.2020 content. The 8-bit path doesnt do that: when colorspace is unspecified, it still branches on color_range and uses BT.601 full-range for AVCOL_RANGE_JPEG.

Should the 16-bit fallback mirror that behavior? As written, a >8-bit full-range or partially tagged source could be silently decoded as limited-range.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants