Duplicated from Plachta/VALL-E-X
b1e1a76
1
2
3
4
5
6
7
8
9
10
11
12
# Text NUM_TEXT_TOKENS = 2048 # Audio NUM_AUDIO_TOKENS = 1024 # EnCodec RVQ bins NUM_MEL_BINS = 100 # BigVGAN bigvgan_24khz_100band # Speaker NUM_SPEAKER_CLASSES = 4096 SPEAKER_EMBEDDING_DIM = 64
