Swallow ZeroGPU low-level CUDA init guard in torch.cuda.set_device (speechbrain calls it unconditionally at init)

Adapt to ZeroGPU: refactor read into @spaces.GPU closure; drop torch/torchvision pins

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

change default loudness

Update app.py

Update app.py

update to the current version

Update app.py

Update app.py

Update app.py

finalize

remove the spaces environment, since we now have a permanent GPU

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

try to figure out how ZeroGPU works

use explicit code instead of relying on release download

use explicit code instead of relying on release download

use explicit code instead of relying on release download

adapt the version number to the release

make read function device agnostic

implement zero GPU compatible solution (hopefully)

initial code release