dall-e-x / models /macros.py
soiz1's picture
Create macros.py
da18641 verified
raw
history blame contribute delete
200 Bytes
# 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