CooperBench/Qwen2.5-Coder-0.5B-Merge-Resolver
Adapter vs Qwen/Qwen2.5-Coder-0.5B-Instruct. CooperBench/Qwen2.5-Coder-0.5B-Merge-Resolver is a LoRA adapter, not full weights: rank 16, on k_proj, o_proj, q_proj, v_proj in 24 layers (96 adapted matrices). It modifies Qwen/Qwen2.5-Coder-0.5B-Instruct at load time.