Skip to content

Dinov2 support on cuda backend#18152

Merged
Gasoonjia merged 7 commits intomainfrom
dinov2-cuda
Mar 15, 2026
Merged

Dinov2 support on cuda backend#18152
Gasoonjia merged 7 commits intomainfrom
dinov2-cuda

Conversation

@Gasoonjia
Copy link
Contributor

@Gasoonjia Gasoonjia commented Mar 13, 2026 •
edited
Loading

Add DINOv2 image classification support for the ExecuTorch CUDA backend.

  • Export: Python script to export facebook/dinov2-small-imagenet1k-1-layer (a lightweight 1-layer DINOv2 ViT-S/14 variant) to .pte + .ptd with CUDA backend, supporting both Linux and Windows targets via --backend cuda|cuda-windows
  • C++ Runner: Image classification runner using stb_image for loading/resizing, ImageNet normalization, and bf16 input/output. Prints top-k predictions with ImageNet class labels
    Build: CMake + CMakePresets + Makefile target (make dinov2-cuda)
  • CI: Integrated into cuda.yml workflow — exports model artifact and runs e2e test with a real dog image, checking for expected output ("Samoyed")

@pytorch-bot
Copy link

pytorch-bot bot commented Mar 13, 2026 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18152

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

❌ 1 New Failure, 1 Cancelled Job

As of commit 48f6862 with merge base 090af6c (image):

NEW FAILURE - The following job has failed:

CANCELLED JOB - The following job was cancelled. Please retry:

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 Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Mar 13, 2026
@github-actions
Copy link

github-actions bot commented Mar 13, 2026

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

@digantdesai
Copy link
Contributor

digantdesai commented Mar 13, 2026

LGTM. Update the title to reflect adding example, and also CPU/XNNPACK tests. Thanks.

@digantdesai digantdesai self-requested a review March 13, 2026 16:06
@mergennachin
Copy link
Contributor

mergennachin commented Mar 13, 2026

Update the PR summary with description, and what's your Test Plan?

@Gasoonjia
Copy link
Contributor Author

Gasoonjia commented Mar 13, 2026

Update the PR summary with description, and what's your Test Plan?

just update ci

@Gasoonjia Gasoonjia temporarily deployed to upload-benchmark-results March 13, 2026 22:18 — with GitHub Actions Inactive
@Gasoonjia Gasoonjia temporarily deployed to upload-benchmark-results March 15, 2026 01:28 — with GitHub Actions Inactive
@Gasoonjia Gasoonjia merged commit abc0237 into main Mar 15, 2026
384 of 386 checks passed
@Gasoonjia Gasoonjia deleted the dinov2-cuda branch March 15, 2026 07:59
jpiat pushed a commit to jpiat/executorch that referenced this pull request Mar 17, 2026
@Gasoonjia @jpiat
Dinov2 support on cuda backend (pytorch#18152)
8f24532 
Add DINOv2 image classification support for the ExecuTorch CUDA backend.

- Export: Python script to export
facebook/dinov2-small-imagenet1k-1-layer (a lightweight 1-layer DINOv2
ViT-S/14 variant) to .pte + .ptd with CUDA backend, supporting both
Linux and Windows targets via --backend cuda|cuda-windows
- C++ Runner: Image classification runner using stb\_image for
loading/resizing, ImageNet normalization, and bf16 input/output. Prints
top-k predictions with ImageNet class labels
Build: CMake + CMakePresets + Makefile target (make dinov2-cuda)
- CI: Integrated into cuda.yml workflow — exports model artifact and
runs e2e test with a real dog image, checking for expected output
("Samoyed")
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@digantdesai digantdesai digantdesai approved these changes

@mergennachin mergennachin mergennachin approved these changes

@larryliu0820 larryliu0820 Awaiting requested review from larryliu0820 larryliu0820 is a code owner

@kirklandsign kirklandsign Awaiting requested review from kirklandsign kirklandsign is a code owner

@lucylq lucylq Awaiting requested review from lucylq lucylq is a code owner

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@Gasoonjia @digantdesai @mergennachin