Refactor block.py #270
Draft
Refactor block.py #270
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
For population connections, these dummy axons were still having an effect. The synapses associated with these axons cause index errors in the NxSDK axon compiler, so those are removed as well. Fixes the "invalid index" error raised in #246.
This allows population spikes to be sent from host to chip when `precompute=True`. Also add spikes to spike generator more quickly by doing deobfuscation once only per group of spikes. Co-authored-by: Eric Hunsberger <[email protected]>
We were accidentally returning the Loihi weights, due to double usage of the term `weights`. This is now less ambiguous.
We avoid the restriction of the axon compartment bases being less than 256, by taking any part of the required offset over this and adding it into the indices instead. Also clarified the `conv2d_loihi_weights` function. Also remove nengo-dl from `test_conv.py:test_conv_connection`. The nengo-dl portions have not been run in a long time, since they were only enabled if `channels_last=True`, and we were x-failing the test for `channels_last=True` due to lack of nengo-loihi support. It contributes nothing to testing nengo-loihi, so remove it.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In #261 I had some questions about why a
spike is Nonecondition was necessary. In investigating that, I started refactoringAxonto reduce the number of ways you could misuse it, as currently it requires you to construct it, then set thetargetoutside of theAxonconstructor, and then callset_compartment_axon_map, and if you don't do one of those steps it will either fail or give you undocumented behavior.While I was able to alleviate those concerns, when I tired to remove the possibility of returning
Noneinstead of a spike it became obvious that several classes insideblock.pyhave similar assumptions, and so several unit tests were failing (despite many integration tests passing). This made me think that there are likely several more issues in how the classes inblock.pyare designed versus how they are actually used in Nengo Loihi, and that seemed outside of the scope of the original issue, so I stopped there. We can pick up this refactoring later if we want.