Average spike rate: 0.0 about snn_toolbox

[rbodo]

Hi,

I can't reproduce the issue; when I run the example (also with your modifications), I get an average spike rate of 0.082...

I'm using tensorflow 2.2.0. I don't think it will make a difference, but you shouldn't explicitly have to install the -gpu version, that is contained in plain pip install tensorflow now. Also, keras does not need to be installed separately any more, the toolbox now uses keras from within tf. Maybe that solves it, otherwise you could try setting a breakpoint here and check whether the spiketrains variable contains any nonzero values.

[dahexiangdongliu]

Hi,

Thanks for your reply! I have created a new environment and installed matplotlib, then the average spike rate was right.

I have set a breakpoint where you recommended, and I found that the self.spiketrains_n_b_l_t was None when the matplotlib was not installed.

When I have installed matplotlib, and removed the config['output'] of the config of mnist_keras_INI.py, the self.spiketrains_n_b_l_t was also None.

Does it mean the spikes printed here is just for plotting, instead of the real spikes when the toolbox is simulating?

[rbodo]

That is correct. When running a large network, recording the spiketrains gets very time and memory consuming, so it is enabled only if needed for plotting. Of course the spikes are still there, we just don't write them out.

In any case, the toolbox shouldn't report a zero rate, I fixed this now.

Thanks for noticing.