Pilotnet-SNN - SpikingDataLoader() output not show as graded or binary - they all show zero values

[gwgkarunasena57]

Dear Experts,

I am running Pilotnet_snn/run.ipynb notebook. I am having following problems related to that could you please help me regarding these questions?

• I loaded some sample data to the SpikingDataLoader as follows.

dataloader = io.dataloader.SpikeDataloader(dataset=full_set,interval=16)

• And try to display the spiking outputs from the SpikingDataLoader() i tried to see them as follows.
  dataloader.data.get(1)

But all the values of the the dataloader's output for that is zero though it shows the shape of the output for that element is as follows as expected.

• So why above all the values are zero? I hope the output should contains binary spikes(zeros as well as ones ) for 16 time steps. Isn't it?

• And if the output of the SpikeDataloader is binary spikes of 16 bit steam that go for 16 time steps, why the Input layer of the pilotnet snn is configure for graded spikes as follows?
  net.in_layer.neuron.du.init = -1

• On the otherhand ,
  - if the input is a graded spike, then why does it need 16 timesteps?
  - if the input is a binary spikes , why does it need input layer, because in that case we can directly feed the binary spikes as in oxford snn example in following link?
  -/dl/netx/oxford/run.ipynb](https://github.com/lava-nc/lava-dl/blob/main/tutorials/lava/lib/dl/netx/oxford/run.ipynb)

  • Beause in that the when i print the relevant SNN it shows as follows, that is without any input layer.

  |   Type   |  W  |  H  |  C  | ker | str | pad | dil | grp |delay|
  

  |Dense | 1| 1| 256| | | | | |False|
  |Dense | 1| 1| 200| | | | | |False|

• And also when i tried to print the input_message_bits for the pilotnet network 'net' as follows it gives the value 0 as output.
  net.input_message_bits

• However in the "PilotNetNxEncoderModel" when the input from lava is adapting to loihi, it uses 16 message bits as follows.

self.adapter = eio.spike.PyToN3ConvAdapter(shape=shape,
num_message_bits=16,
compression=compression)
self.conv = Conv(weight=np.eye(3).reshape(3, 1, 1, 3),
input_shape=shape, num_message_bits=16)

Thank you.

Thanks and Rgds,
Udayanga