antoninbrthn/skill-neologisms-qwen2.5-digitseq-base
Adapter vs Qwen/Qwen2.5-0.5B. antoninbrthn/skill-neologisms-qwen2.5-digitseq-base is a LoRA adapter, not full weights: rank 32, on down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj in 24 layers (168 adapted matrices). It modifies Qwen/Qwen2.5-0.5B at load time.