The commit above adds a new test script for the delays, unfortunately the performance is pretty awful... @Hananel-Hazan If you'd like to take a look at it feel free. It's essentially just a copy of the basic reservoir.py script with layers that feature the new topology feature.

We can omit it if you want, but probably good to have an example that shows it's at least reasonably functional.

Script output and Performance:

Running on Device =  cpu
Train progress: (500 / 500): 501it [11:45,  1.41s/it]
Data extraction progress: (500 / 500): 501it [12:03,  1.44s/it]

 Training the read out
Epoch: 500/500, Loss: 2.5968: 500it [00:23, 21.61it/s]
Testing progress: (500 / 500): 501it [12:23,  1.48s/it]

 Accuracy of the model on 500 test images: 9.18 %