search

NeuroBench / neurobench

Benchmark harness and baseline results for the NeuroBench algorithm track.
https://neurobench.readthedocs.io
Apache License 2.0
46 stars 11 forks source link

S2S first timestep removal #193

Closed jasonlyik closed 3 months ago

jasonlyik commented 4 months ago

The S2S algorithm seems to have timesteps = (sample_frequency / hop_length) + 1. Is this intended behavior? We could remove the first timestep by adding spikes = spikes[:,1:,:] before returning.

jasonlyik commented 3 months ago

TODO: check accuracy on ANN/SNN benchmark when reduced to 200 timesteps

jasonlyik commented 3 months ago

Seems like there is negligible change in accuracy / sparsity by removing the first timestep on both the GSC and the MSWC tasks.