Closed arvoelke closed 2 years ago
It looks like somewhere we're assuming that each neuron has at least one input in the connection. As a workaround, you can add extra values so that each neuron has at least one input synapse (even if the weight is 0).
Looks like that works, although instead of exactly 0
it should be something small like 1e-20
-- zeros must be getting removed somewhere. Thanks!
n_neurons = 1000
edge_list = []
weights = []
edge_list.append((0, 0))
weights.append(1)
for i in range(1, n_neurons):
edge_list.append((0, i))
weights.append(1e-20)
transform = Sparse((n_neurons, n_neurons), edge_list, weights)
Fixed in #246.
I no longer get this error on master
, so closing.
This occurs for both the emulator and the actual hardware. If I change
n_neurons
to1
I get a different error:And if I change the simulator to
nengo.Simulator
then both work. This is usingnengo==3.0.0.dev0
andnengo_loihi==0.9.0.dev0
.