Files
2026-04-07 21:22:30 +00:00

16 lines
639 B
YAML

CODEC_TYPE: "snac" # "snac" or "mimi"
CODEC_MODEL: null # null = default model per codec type
NUM_CODEBOOKS: 8 # mimi: 1-32 (default 8), snac: ignored
TOKENIZER_LENGTH: 128256
START_OF_TEXT: 128000
END_OF_TEXT: 128009
START_OF_SPEECH: 128257 # TOKENIZER_LENGTH + 1
END_OF_SPEECH: 128258 # TOKENIZER_LENGTH + 2
START_OF_HUMAN: 128259 # TOKENIZER_LENGTH + 3
END_OF_HUMAN: 128260 # TOKENIZER_LENGTH + 4
START_OF_AI: 128261 # TOKENIZER_LENGTH + 5
END_OF_AI: 128262 # TOKENIZER_LENGTH + 6
PAD_TOKEN: 128263 # TOKENIZER_LENGTH + 7
AUDIO_TOKENS_START: 128266 # TOKENIZER_LENGTH + 10