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Outdated suggestions cannot be applied.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.
When implementing batching I wrote a helper function called
genVariableFill
for initialising variables, taking account of delays and batching. However, for some reason, I forgot to actually use this when initializinginSyn
,denDelay
(which was subsequently copy-pasted into the initialisation ofrevInSyn
). CUDA seems to zero newly allocated memory so, generally, this has minimal effect but, if you reinitialise your model, you may be left with large residual neural input currents in some batch instances (which, if you're using dendritic delays will continue to appear for max dendritic delay timesteps!).This finally fixes the issues we were seeing in Felix Schmitt's model - big thanks to him for testing my various speculative fixes and suggesting that (re-)initialization could be the problem!