anonymous-submission-ICLR2027/Gobble-Prover-1.7B
LoRA-style vs AI-MO/Kimina-Prover-Distill-1.7B. 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 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/113, MLP 84/84, Attention 112/112
1 derivatives checked
- mradermacher/Gobble-Prover-1.7B-GGUF: Same weights, converted