New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Windows Machine Learning Code Generator Not Working #290
Comments
Hi @Janickvw |
Hi @LPBourret. I have tried the extension "Windows Machine Learning Code Generator" as per the title in both VS 2017 and 2019 with no luck. Even tried uninstalling the extension and reinstalling on both. Also tried uninstalling VS 2019 and reinstalling. No luck there either. |
Hi @Janickvw |
@LPBourret Hi yes, I tried excluding and re adding the model many times after the extension was installed. Also tried excluding the model, closing VS, reopening and re adding to no luck. |
I am seeing the same exact behavior in VS2019. I have tried install, uninstall/reinstall, disable/enable, different models from known UWP sample apps, model from Custom Vision - nothing happens when adding an onnx file to the project. |
Hi @zhangxiang1993 this is the model https://gallery.azure.ai/Model/MNIST-Handwritten-Digit-Recognition . I assumed this model would work with MLgen as it is literally the model used in the MLgen tutorial demo, so it doesn't make sense this wouldn't work. Also tried using my custom vision object detection model which didn't work either. |
@Janickvw Weird, for me with Visual Studio 2019 (v16.4.2) and this Windows Machine Learning Code Generator plugin, the model you linked leads to correct automatically generated wrapping code. The model is of an older ONNX opset that was supported by CNTK 2.4 so the ONNX opset version is not the issue here. |
Just applied the 1/2/2020 update to WMLCG and still no joy. Very disappointing. No errors. Muddling through how to build the c# wrapper using online code examples. |
MLGEN is not working ! I tried with different versions under 1.3 and still no .cs generated. |
It may be a path character problem. For example, code generation failed in c:\한글\WIndows-Machine-Learning. It created the mnist.cs file to Assets folder but no code in it. So I moved the Sample folder to c:\WIndows-Machine-Learning and it works. |
I'm experiencing the same issue in Visual Studio 2019 with an onnx model version 1.5 and 1.3. I already installed "Windows Machine Learning Code Generator", reopened VS, deleted my project and tried with a new one. nothing worked |
@miyrai , would you be able to share your model for investigation purposes? I was able to get onnx models version 1.3 and 1.5 working on my end. |
@ryanlai2 I used this one https://github.com/onnx/models/blob/master/vision/object_detection_segmentation/mask-rcnn/model/MaskRCNN-10.onnx and i also tried your suggested models, but it didn't work with both of them. I know that the problem is with my settings, because it works for you, so what i did is:
|
@ryanlai2 if it works for you that the cs file of my onnx model is generated, maybe you could just send me the cs file? thats all i need |
Here you go @miyrai ! Let me know if it works :) You will need to change the namespace. |
It works |
For anyone still running into problems win the mlgen: The extension doesn't work if the full path to your project contains white spaces! |
I'm submitting a…
Bug report (I searched for similar issues and did not find one)
Current behavior
When adding an onnx model with right click on assets -> add existing item . No code is generated but onnx model appears in assets folder.
Expected behavior
Interface file to be generated.
Minimal reproduction of the problem with instructions
Repo - Stock MNIST tutorial demo.
Steps:
Environment
Windows Build Number: 18362.535
App min and target version: 17763, 18362
OS Version (Server, IoT Core, Desktop, etc): Windows 10 Pro
Graphics Driver version: 25.20.100.6472
DxDiag: 10.00.18362.0387
WinMLTools specific:
- Source training framework: (e.g. CoreML, Scikit-learn, …)
- WinMLTools version
Visual Studio
tried with both
The text was updated successfully, but these errors were encountered: