Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Remove (N, T, H, W, C) => (N, T, C, H, W) from presets #6058

Merged
merged 7 commits into from
May 23, 2022
Merged

Remove (N, T, H, W, C) => (N, T, C, H, W) from presets #6058

merged 7 commits into from
May 23, 2022

Conversation

datumbox
Copy link
Contributor

@datumbox datumbox commented May 20, 2022 •
edited
Loading

Should be merged after #5952.

In this PR we:

  • Remove the (N, T, H, W, C) => (N, T, C, H, W) conversion from the presets
  • Update the example on the docs to explicitly permute the output of read_video()
  • Add output_format param in Kinetics dataset and read_video() usages.
  • Add better descriptions on presets to describe all accepted inputs and their dimensions.
  • Verify that the accuracy of pre-trained models using the reference scripts

The accuracy looks similar to #5618. All good:

torchrun --nproc_per_node=8 train.py --data-path /datasets01/kinetics/070618/400/ --batch-size=16 --sync-bn --test-only --weights R2Plus1D_18_Weights.KINETICS400_V1 --cache-dataset
* Clip Acc@1 56.496 Clip Acc@5 77.756

@datumbox datumbox mentioned this pull request May 20, 2022
Merged
@datumbox datumbox requested a review from NicolasHug May 23, 2022 12:09
NicolasHug
NicolasHug approved these changes May 23, 2022
View reviewed changes
Copy link
Member

@NicolasHug NicolasHug left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @datumbox

@datumbox
Copy link
Contributor Author

The macOS and ONNX jobs are broken and unrelated to this PR. Merging as discussed with @NicolasHug offline.

@datumbox datumbox merged commit 60ce5bf into pytorch:main May 23, 2022
@datumbox datumbox deleted the models/use_nchw branch May 23, 2022 12:55
facebook-github-bot pushed a commit that referenced this pull request Jun 1, 2022
@YosuaMichael @facebook-github-bot
[fbsync] Remove (N, T, H, W, C) => (N, T, C, H, W) from presets (#6058
e261ab6 
)

Summary:
* Remove `(N, T, H, W, C) => (N, T, C, H, W)` conversion on presets

* Update docs.

* Fix the tests

* Use `output_format` for `read_video()`

* Use `output_format` for `Kinetics()`

* Adding input descriptions on presets

Reviewed By: NicolasHug

Differential Revision: D36760943

fbshipit-source-id: 316f98583f39cc29b9a40f9c7c479b565981f088
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@NicolasHug NicolasHug NicolasHug approved these changes

Assignees
No one assigned
Labels
cla signed enhancement module: models
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@datumbox @NicolasHug @facebook-github-bot