jeff-legacy/Jeff-Qwen3.5-0.8B-ground
Adapter vs jeff-legacy/Jeff-Qwen3.5-0.8B. jeff-legacy/Jeff-Qwen3.5-0.8B-ground is a LoRA adapter, not full weights: rank 16, on down_proj, gate_proj, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj… in 24 layers (150 adapted matrices). It modifies jeff-legacy/Jeff-Qwen3.5-0.8B at load time.