Why is MBD not allowed for stereo?

[ajayarora1235]

They specify how to do MBD for stereo in the docs, here's how I've implemented it. Happy to make a PR for this as well.

output_2 = model_2.generate(
    descriptions=[
        description
    ],
    progress=True, return_tokens=True
)

tokens = output_2[1]
left, right = model_2.compression_model.get_left_right_codes(tokens)
tokens = torch.cat([left, right])
outputs_diffusion_2 = mbd.tokens_to_wav(tokens)
assert outputs_diffusion_2.shape[1] == 1  # output is mono
outputs_diffusion_2 = rearrange(outputs_diffusion_2, '(s b) c t -> b (s c) t', s=2)
``` /