thjashin / multires-conv

Sequence Modeling with Multiresolution Convolutional Memory (ICML 2023)
MIT License
119 stars 2 forks source link

drop in replacement for conv1D? #4

Closed listener17 closed 6 months ago

listener17 commented 7 months ago

Thanks for sharing this interesting work! Can the MultiresConv be utilized as a drop-in replacement for conv1d operating on audio signals?

Also, is it possible to build an inverse or transpose MultiresConv and use it as a drop-in replacement for transposed conv1d?

thjashin commented 7 months ago

Thanks for your interest! Short answer to the first question: Yes. Think of it as a general sequence-to-sequence mapping. The inverse MultiresConv idea is interesting. I haven't thought about it deeply but it looks plausible.

listener17 commented 6 months ago

Thanks @thjashin

listener17 commented 5 months ago

@thjashin: if I would like to replace the regular conv1d layer in my model with your multires layer, how should I call your layer?