Why I can not convert my ANN model to SNN successfully?

NeuromorphicProcessorProject / snn_toolbox

Toolbox for converting analog to spiking neural networks (ANN to SNN), and running them in a spiking neuron simulator.

MIT License

362 stars 105 forks source link

Why I can not convert my ANN model to SNN successfully? #10

Closed oxygenman closed 7 years ago

oxygenman commented 7 years ago

The attach file is the model of ANN I use. The architecture of the model is shown below. But when I performing the converting program, it failed with the err： model

model.zip

It seems my model have an "InputLayer" that the toolbox can't understand, how can I solve this problem?

rbodo commented 7 years ago

Fixed this bug (it was due to the branching directly after the input layer).

However, there seems to be a problem with the final MaxPool layer (the other Pooling layers work fine). Replacing Max by AveragePool also works. I will get back to you with a solution hopefully soon.

oxygenman commented 7 years ago

OK，thank you very much, I will keep learning your code!

rbodo commented 7 years ago

OK, all good now. The problem was using data_format 'channels_last' instead of 'channels_first'. Both are now supported. I tested your model with random input and it ran fine.