Prediction using trained model

[Thelbock]

Hello @pradeep-pyro,

first of all, thank you very much for sharing the amazing work of you and your team!

I managed to train the Segmentation Model using my own labeled step files.
The results are very good and now I want to do predictions on single, unseen graph files.

I pass a batched DGL graph to the 'forward' method of the model, but the problem is that I always get an 'Assertion Error'.
Here you can see some tries I did, but all of them result in the same error:

At the beginning I loaded the model by using 'Segmentation.load_from_checkpoint(...).

I would be very thankful for your help or maybe a simple code snipped of how to do predictions on a single graph file.

Thank you very much,

Tim

[pradeep-pyro]

Hi @Thelbock, can you share details on the assertion error?

[Thelbock]

Hi @pradeep-pyro and thank you for your answer.

When I debug my code (e.g. for line 145 in the previous image), I get the assertion error when the forward method of the models.py class is called. As you can see in the following image, line 247 outputs the assertion error:

Unfortunately there is not more information about the error.

Here is the .bin file I wanted to use for the prediction, and the original step file from which the .bin file was created.
uvnet_prediction.zip

Thanks again,

Tim

[Fdioa]

hi, @Thelbock Have you found the right way to make predictions using a single bin file?