Add GPU inference test and CI workflow integration

[Borda]

This pull request introduces a new GPU-specific continuous integration (CI) workflow and adds a comprehensive GPU inference benchmark test for COCO detection models. The main goal is to ensure that GPU-enabled tests are automatically run on relevant branches and pull requests, and that the core detection models meet minimum accuracy thresholds on the COCO dataset.

CI/CD Improvements:

• Added a GitHub Actions workflow (.github/workflows/ci-tests-gpu.yml) to automatically run GPU-based tests on pushes and pull requests targeting the main and develop branches. This workflow sets up a GPU environment, installs dependencies, runs tests marked for GPU, collects coverage, and uploads results to Codecov.

Testing Enhancements:

• Introduced a new test suite (tests/test_inference_gpu.py) that benchmarks the RFDETRXLarge and RFDETR2XLarge models on the COCO validation dataset using GPU. The test checks that each model meets specified minimum mAP@50 and F1-score thresholds, and dumps evaluation statistics for debugging.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89%. Comparing base (57ddaa8) to head (1a59874).

Additional details and impacted files

@@        Coverage Diff         @@
##           main    #5   +/-   ##
==================================
  Coverage    89%   89%           
==================================
  Files         4     4           
  Lines        62    62           
==================================
  Hits         55    55           
  Misses        7     7           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This pull request adds GPU-specific continuous integration infrastructure and a comprehensive GPU inference benchmark test for RF-DETR+ models. The PR extends the existing CPU-only testing setup to include GPU-based validation of the XLarge and 2XLarge model variants against COCO dataset accuracy thresholds.

Changes:

• Added a GitHub Actions workflow (.github/workflows/ci-tests-gpu.yml) to run GPU-marked tests on a custom GPU runner
• Introduced a parametrized benchmark test (tests/test_inference_gpu.py) that validates XLarge and 2XLarge models meet minimum mAP@50 and F1-score thresholds on COCO validation data

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 10 comments.

File	Description
.github/workflows/ci-tests-gpu.yml	Defines GitHub Actions workflow to run GPU tests on Roboflow-GPU-VM-Runner with nvidia-smi validation, pytest execution, and Codecov reporting
tests/test_inference_gpu.py	Implements parametrized GPU benchmark test for XLarge and 2XLarge models using COCO validation dataset with configurable accuracy thresholds

[Copilot]

The download_coco_val fixture is used but not defined anywhere in the test suite. No conftest.py file exists in the tests directory that would define this fixture. This will cause the test to fail immediately when pytest tries to collect it. You need to either create a conftest.py file with this fixture defined, or define it directly in this test file.

[Borda]

Yeah, need to port it