Add YOLO model family check with ONNX import and test #3750
+252
−73
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Introduces a new model-checks/yolov8n crate for Burn, including Cargo.toml, build script for ONNX codegen, Python script to download and process the YOLOv8n model and generate test data, and a Rust main.rs to validate model output against reference data. This enables automated verification of YOLOv8n ONNX import and inference correctness across supported backends.
|
Running with metal backend fails. CC @wingertge |
24 tasks
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #3750 +/- ##
=======================================
Coverage 64.01% 64.01%
=======================================
Files 1084 1084
Lines 126880 126880
=======================================
Hits 81228 81228
Misses 45652 45652 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
That looks like a fusion error, so I think @nathanielsimard would know more here |
|
Actually it could also be an issue fixed by tracel-ai/cubecl#895 |
Renamed yolov8n model-checks crate to yolo and refactored code to support multiple YOLO variants (yolov5s, yolov8n, yolov8s, yolov10n, yolo11x). Added model selection via YOLO_MODEL environment variable, updated build script, Python model preparation script, and main.rs to handle dynamic model selection and output. Added README with usage instructions and supported models. Removed yolov8n-specific files.
Deleted the YOLO11x model check directory, including Cargo.toml, build script, model preparation Python script, and main Rust source. This removes support and tests for the YOLO11x model from burn-import/model-checks.
antimora
changed the title
I'm not sure. I'm running with that latest code and still seeing the issue with my YOLOv9m model when using Wgpu and Metal (but not with just Wgpu). |
laggui
approved these changes
Sep 29, 2025
There was a problem hiding this comment.
Thought this was being held up by the test failure, but actually this is captured in #3780 (comment).
LGTM!
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Introduces a model verification system for YOLO family models including yolov5s, yolov8n, yolov8s, yolov10n, and yolo11x. This PR includes a new model-checks crate for Burn with Cargo.toml, build script for ONNX codegen, Python script to download and process YOLO models and generate test data, and Rust implementation to validate model outputs against reference data. This enables automated verification of YOLO model ONNX import and inference correctness across supported backends.
Pull Request Template
Checklist
cargo run-checkscommand has been executed.Related Issues/PRs
#2822
Changes
Added model-checks crate with support for YOLO family models (yolov5s, yolov8n, yolov8s, yolov10n, yolo11x)
Testing
Successfully runs for tch and ndarray backend but fails for metal backend. yolov10n also fails with TopK and Mod ONNX ops. PRs to fix the ops will follow.