search

JusperLee / Dual-Path-RNN-Pytorch

Dual-path RNN: efficient long sequence modeling for time-domain single-channel speech separation implemented by Pytorch
Apache License 2.0
417 stars 65 forks source link

multichannel #19

Open shanhaidexiamo opened 4 years ago

shanhaidexiamo commented 4 years ago

Hi, I am a newer in speech separation. I'm trying to do multichannel SS in time domain, and I want to ask if I do the multichannel, what things should I pay attention to in DPRNN?

JusperLee commented 4 years ago

It is recommended that you take a look at this method to handle multi-channel speech separation： https://github.com/yluo42/TAC

shanhaidexiamo commented 4 years ago

ok! thank you very much!

shanhaidexiamo commented 4 years ago

it seems that if batchsize is set to 2 ,there will be a runtime error about cuda is out of memory, is it right?

JusperLee commented 4 years ago

Generally speaking, when batch size=1, the occupied Cuda memory is about 9000M.

shanhaidexiamo commented 4 years ago

OK , I got it. Thank you again for your reply.