Getting KeyError: 'dep_q' while using moshi
#45
by
Swarn - opened
Getting KeyError: 'dep_q' while running the server or in -mimi = checkpoint.get_mimi(device="cuda")
Is there some incompatibility between moshi and personaplex?
Yes there are incompatibilities. 1) Moshi does not support voice/text prompt section. 2) Moshi depformer has 8 channels of audio, while PersonaPlex has 16 due to historical trainings reasons which is no longer relevant but needs to be cleaned up. So PersonaPlex will likely not work with Moshi inference code, and Moshi will likely not work with PersonaPlex inference code.
royrajarshi changed discussion status to
closed