TiGa-RCE/Bonsai-27B-MLX-BF16-Config-Repaired

Same weights, converted vs prism-ml/Bonsai-27B-gguf. Same weights stored in another dtype (F16/F32 → BF16), plus 48 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/161, MLP 0/192, Attention 0/64

2 derivatives checked