Hanno-Labs/dinghy-law-8b-v1
LoRA-style vs nvidia/Nemotron-3-Embed-8B-BF16. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 34 of 34 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/69, MLP 102/102, Attention 136/136, Embeddings 0/1
1 derivatives checked
- Hanno-Labs/dinghy-law-8b-v1-gguf: Same weights, converted