"Uncaught (in promise) TypeError: _this3.signature.inputs is undefined" when trying to predict from UltraLytics YOLOv5 implementation when exported to TFJS

[NormanTUD]

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow.js): Yes
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Debian 11
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device: No
• TensorFlow.js installed from (npm or script link): No
• TensorFlow.js version (use command below): 4.0.0
• Browser version: Firefox 102, Chromium 106.0.5249.91
• Tensorflow.js Converter Version: 3.15.0

Describe the current behavior

When I'm trying to predict an exported model from UltraLytics YOLOv5 Implementation, which has been exported with python3 export.py --weights runs/train/exp/weights/best.pt --img 640 640 --batch-size 1 --include tfjs and loaded with var model = await tf.loadGraphModel('./model.json') (prediction via var res = await model.executeAsync(img_tensor);), I get the following error message and the program stops execution.

Describe the expected behavior

I expected the model to "just work".

Other info / logs Include any logs or source code that would be helpful to
diagnose the problem. If including tracebacks, please include the full
traceback. Large logs and files should be attached.

This problem was fixed by changing line 93481 in tf.js (from https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@4.0.0/dist/tf.js) from

var signature = _this3.signature ? _this3.signature.inputs[inputName] : null;

to

var signature = Object.keys(_this3).includes("signature") && Object.keys(_this3.signature).includes("inputs") ? _this3.signature.inputs[inputName] : null; 

I have no idea why it fails or where exactly to change this in the "real" TypeScript (not the JS only), and for me this quickfix is enough, since my project works with it and is not meant for external users currently.

But it would be good if it was fixed in the proper version, too.

[pyu10055]

@NormanTUD Thank your for reporting, this has been fixed with this PR.
It should be available in the next release for TFJS.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No