mlx-community/OmniVoice-bfloat16
Same weights, converted vs k2-fsa/OmniVoice. Same weights stored in another dtype (F32/I64 → BF16/I64).
Tensors changed: other 0/2, Embeddings 0/2, Norms 0/113, MLP 0/84, Attention 0/112
Same weights, converted vs k2-fsa/OmniVoice. Same weights stored in another dtype (F32/I64 → BF16/I64).
Tensors changed: other 0/2, Embeddings 0/2, Norms 0/113, MLP 0/84, Attention 0/112