lif subthr_dynamics() for 'fixed_pt' doesn't properly wrap neuron current

[JuliaA369]

Objective of issue: Fix subthr_dynamics() function for fixed point lif model so that neuron current wraps to opposite side of integer range rather than to 0.

Lava version:

• 0.4.0 (current version)
• 0.3.0

I'm submitting a ...

• [ x] bug report

Current behavior:

• Right now on integer overflow of current, the current wraps back around to 0

Expected behavior:

• You'd expect the current to wrap around to -2^23 if the current is positively accumulating or +2^23 if accumulating negatively. The comments imply that this is also the expected behavior

Steps to reproduce:

• Set current decay to be really slow, and have a bunch of spiking inputs go into a single neuron, then probe current and look at it on the plot

Related code:

insert short code snippets here

Other information:

A simple way to fix this issue is to go into lava/proc/lif/models.py, then find the fixed point version of subthr_dynamics() and change the wrapped_curr computation to the following:

wrapped_curr = np.where(decayed_curr>self.max_uv_val, decayed_curr - 2*self.max_uv_val,
                 decayed_curr)
wrapped_curr = np.where(wrapped_curr<=-self.max_uv_val, decayed_curr + 2*self.max_uv_val,
                 wrapped_curr)