Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
Tensor<Spikes> -> CompressedTensor<Spikes>
Compress binary spike tensors during the forward pass and decompress them dynamically during backpropagation.
Problem it solves
Backpropagation-through-time caches state sequences across all steps, exhausting GPU memory.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.