t0rr3sp3dr0/Qwen3.8-27B-MLX-MTP-mxfp8
Partial fine-tune (fewer layers) vs mlx-community/Qwen3.8-27B-MTP-mxfp8. mlx-community/Qwen3.8-27B-MTP-mxfp8 matches part of t0rr3sp3dr0/Qwen3.8-27B-MLX-MTP-mxfp8 (21 of its 1703 tensors; the rest of t0rr3sp3dr0/Qwen3.8-27B-MLX-MTP-mxfp8 has no counterpart). On the shared part: Partial change: 0% of the weights changed (embeddings, norms), in 1 of 1 layers.
Tensors changed: Norms 6/6, MLP 0/6, Attention 0/8, Embeddings 1/1