# original BARK semantic vocab size
SEMANTIC_VOCAB_SIZE = 10_000
# HuBERT model output vocab size
HUBERT_OUTPUT_VOCAB_SIZE = 10_003
CODEBOOK_SIZE = 1024
N_COARSE_CODEBOOKS = 2
COARSE_RATE_HZ = 75
COARSE_SEMANTIC_PAD_TOKEN = 12_048
COARSE_INFER_TOKEN = 12_050

# for the BERT model to get semantic tokens from raw texts
TEXT_ENCODING_OFFSET = 10_048
SEMANTIC_PAD_TOKEN = 10_000
TEXT_PAD_TOKEN = 129_595
SEMANTIC_INFER_TOKEN = 129_599
SEMANTIC_RATE_HZ = 49.9

N_FINE_CODEBOOKS = 8