Joseph Turian

Particularly demonstrating a multi-GPU setup

In the VCOs, we still use make_control_as_frequency. We don't need to upsample midi_f0 several times, I think we can use the UpsampleModule

``` from torchsynth.synth import Voice gpuvoice = Voice() gpuvoice.eval() gpuvoice.to("cuda") gpuvoice.on_post_move_to_device() ``` Do we have to use on_post_move_to_device()?

DEBUG = True in tests, but not in prod

1

We want DEBUG = True for CI but False by default for users

Config parameters

1

Allow people easily to specify config parameters when creating Voice

https://github.com/turian/torchsynth/pull/280 See also #140

https://github.com/turian/torchsynth/pull/280 See also #104