Daniel Rasmussen

Hello @kebitmatf, I haven't seen that behaviour before, do you have a minimal example to help debug the issue?

I tried running a memory profiler on your code, but I wasn't seeing the effect you describe (in fact, `predict` was using less memory than `evaluate`). However, one thing to...

> 1. Attaching probe after each layer will lead to full memory This is expected. The probe data needs to be stored in memory, so the more probes you have...

Hi @julvanu, thanks for the bug fix! Could you add a test case for converting a concatenate layer with a single input? You can see the current concatenate test here...

In #142 we updated this to give a more informative error message, but merging Distributions is still not supported.

I think if you do `print(subnet.config[subnet.ens].on_chip)` it will work as expected (because that's where you set it, not `net.config`).

Ah yeah, related to https://github.com/nengo/nengo/issues/1601, basically NengoLoihi only looks at the top-level config, it doesn't check nested configs at all I don't think.

I know there's been a lot of work on the keyword spotting demo since this came up, do you know if this is still an issue @hunse?

Yeah that sounds like a good plan, I'll add it to the backlog for a future sprint.

As pointed out by @hunse, this is caused by the change where we run `connect` in `__enter__` (NxSDK doesn't like it if you try to add energy probes after `connect`).