CUDA Out of memory while trying to use the pre-trained model about vocal-remover

[TRvlvr]

I had the same issue when I tried running this on my old gaming laptop that also had a Nvidia GPU with 2GB's of RAM (bought it brand new in 2014). PyTorch and the newer Cuda drivers it comes with don't play well at all with older GPU's that have anything less than 4GB's of V-RAM.

[tsurumeso]

python inference.py -i file.mp3 -g 0 -w 256 but I got the error

Try setting --window_size (-w) to 384 because it must be greater than 256.

I'm trying to reduce memory usage recently, so it may be fixed in the next version.

[AlbyTree]

Thank you, now it works.

[FlashlightET]

train.py: error: unrecognized arguments: -w 384