prithivMLmods/OneDecision-VisionGuard-27B-SFT
LoRA-style vs Qwen/Qwen3.8-27B. 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, linear_fc1, linear_fc2, o_proj, out_proj, proj, q_proj, qkv, up_proj, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/319, Attention 358/556, MLP 246/304, Embeddings 0/4, lm_head 0/1
2 derivatives checked
- mradermacher/OneDecision-VisionGuard-27B-SFT-GGUF: Quantization
- prithivMLmods/OneDecision-VisionGuard-27B-SFT-GGUF: Same weights, converted