Mr-Shmoo/sonda-1.1-4B
LoRA-style (fewer layers) vs Qwen/Qwen3.5-4B-Base. Mr-Shmoo/sonda-1.1-4B matches part of Qwen/Qwen3.5-4B-Base (426 of its 738 tensors; the rest of Qwen/Qwen3.5-4B-Base has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, MLP 0/96, Attention 152/224, Norms 0/105
1 derivatives checked
- mradermacher/sonda-1.1-4B-GGUF: Same weights, converted