Support Sparse transforms in Simulator.get_nengo_params #149
Conversation
|
|
||
| with Simulator(net) as sim0: | ||
| params = sim0.get_nengo_params(conn) | ||
| assert np.allclose(params["transform"].init.data, data) |
There was a problem hiding this comment.
Will this work if Scipy is not installed? In that case, what will params["transform"].init be? I'm assuming a Numpy array, but right now when I try without Scipy I'm getting Aborted (core dumped). I'll look into it a bit more. Is this expected to work without Scipy, or does NengoDL only support sparse with Scipy (in which case, we should mark this whole test as requiring Scipy)?
There was a problem hiding this comment.
This test should be skipped if scipy isn't installed (in the use_scipy=True condition). In the use_scipy=False condition, .init.data will be SparseMatrix.data.
There was a problem hiding this comment.
I forgot we had the SparseMatrix. Anyway, I reinstalled and uninstalled Scipy and the weird error disappeared, so who knows? But it seems good now.
This functionality (and
Simulator.freeze_params, which depends on it) had not been updated after the introduction of Sparse transforms.Fixes #148