dual_ar_2_codebook_small.yaml 408 B

12345678910111213
  1. _target_: fish_speech.models.text2semantic.llama.DualARTransformer
  2. config:
  3. _target_: fish_speech.models.text2semantic.llama.DualARModelArgs
  4. max_seq_len: ${max_length}
  5. vocab_size: 264 # pad 262 to 8x
  6. n_layer: 12
  7. n_fast_layer: 4
  8. n_head: 12
  9. dim: 768
  10. rope_base: 10000
  11. norm_eps: 1e-5
  12. num_codebooks: 2 # input/output codebook size
  13. codebook_size: 1032 # codebook size 1024 + 2 special tokens