DoHorizon/MarChat-V1-27B
LoRA-style vs Qwen/Qwen3.5-27B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, MLP 192/304, Attention 304/556, Norms 0/319
2 derivatives checked
- mradermacher/MarChat-V1-27B-i1-GGUF: Quantization
- mradermacher/MarChat-V1-27B-GGUF: Quantization