- _target_: fish_speech.models.text2semantic.llama.NaiveTransformer
- config:
- _target_: fish_speech.models.text2semantic.llama.DualARModelArgs
- max_seq_len: ${max_length}
- vocab_size: 36408
- n_layer: 12
- n_head: 12
- dim: 768
- rope_base: 10000
- norm_eps: 1e-5
- num_codebooks: 8 # input/output codebook size
- codebook_size: 264 # codebook size 256 + 2 special tokens
|
