update detect.py in order to support torch script

[andreiionutdamian]

This change assumes the torchscrip file was previously saved with export.py.
Based on the issue #5070

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Added TorchScript compatibility to model loading in detect.py.

📊 Key Changes

• detect.py now includes a conditional to check for 'torchscript' in the model file name.
• If 'torchscript' is found, the script uses torch.jit.load() to load the model; otherwise, it uses the existing attempt_load() method.

🎯 Purpose & Impact

• Purpose: This change enables the use of TorchScript models, which can offer better performance and easier deployments to different platforms.
• Impact: Users can now run inference with YOLOv5 using TorchScript format models, potentially increasing the use cases and environments where YOLOv5 can be applied. 🚀

[github-actions]

👋 Hello @andreiionutdamian, thank you for submitting a 🚀 PR! To allow your work to be integrated as seamlessly as possible, we advise you to:

• ✅ Verify your PR is up-to-date with origin/master. If your PR is behind origin/master an automatic GitHub actions rebase may be attempted by including the /rebase command in a comment body, or by running the following code, replacing 'feature' with the name of your local branch:

git remote add upstream https://github.com/ultralytics/yolov5.git
git fetch upstream
git checkout feature  # <----- replace 'feature' with local branch name
git merge upstream/master
git push -u origin -f

• ✅ Verify all Continuous Integration (CI) checks are passing.
• ✅ Reduce changes to the absolute minimum required for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." -Bruce Lee

[glenn-jocher]

@andreiionutdamian thanks for the PR! I think this covers model loading, but we need additions in a second, place, inference, right? Otherwise you are attempting to use a torchscript model for normal pytorch inference, which should not be possible (probably the source of the CI fails):

yolov5/detect.py

Lines 142 to 145 in 276b674

	# Inference
	if pt:
	visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
	pred = model(img, augment=augment, visualize=visualize)[0]

[andreiionutdamian]

@glenn-jocher it should work just fine. The whole intuition is to have exactly the same functionality but in a more graph/serialized way. Tried this already on other models and it works quite nice. So basically what happens after th.jit.load or th.load (or th.nn.Module definition, th.nn.Module instance & param dict loading) should not matter...

And one more thing: thank you sir for all your replies and real hands-on involvement.

[glenn-jocher]

@andreiionutdamian oh, got it! Good to know.

Can you debug the updates then to resolve the CI failures? Then we should be all set. Thanks!

[andreiionutdamian]

@andreiionutdamian oh, got it! Good to know.

Can you debug the updates then to resolve the CI failures? Then we should be all set. Thanks!

Sorry, forgot to merge the modifications of the sloppy code spacing done directly in GitHub editor.