{ "schema": "dynquant_allocation_v1", "dynquant_core": "0.5.2", "model": "/workspace/runs-s2/qwen38-27b.text2sql/merged", "stats": "/workspace/runs-s2/qwen38-27b.text2sql/stats/dynquant_stats.json", "allocator": "rank_product", "group_size": 128, "maps": { "4.00": { "target_label": "4.00 avg bits", "average_bits": 3.999888904217332, "nbytes": 13447625728, "group_size": 128, "pricing": { "measured_modules": 0, "proxied_modules": 498, "measured_params": 0, "proxied_params": 26893352960, "proxied_share": 1.0, "scale": 1.0 }, "histogram": { "2": 1, "3": 135, "4": 265, "8": 97 }, "violations": [ { "name": "model.layers.0.mlp.gate_proj", "role": "mlp.gate", "floor_bits": 4, "assigned_bits": 3, "num_params": 89128960 }, { "name": "model.layers.0.mlp.up_proj", "role": "mlp.up", "floor_bits": 3, "assigned_bits": 2, "num_params": 89128960 }, { "name": "model.layers.54.linear_attn.in_proj_qkv", "role": "lin_attn.qkv", "floor_bits": 4, "assigned_bits": 3, "num_params": 52428800 }, { "name": "model.layers.55.self_attn.o_proj", "role": "attn.o", "floor_bits": 4, "assigned_bits": 3, "num_params": 31457280 }, { "name": "model.layers.55.self_attn.q_proj", "role": "attn.q_gate", "floor_bits": 4, "assigned_bits": 3, "num_params": 62914560 }, { "name": "model.layers.58.linear_attn.in_proj_qkv", "role": "lin_attn.qkv", "floor_bits": 4, "assigned_bits": 3, "num_params": 52428800 }, { "name": "model.layers.58.linear_attn.in_proj_z", "role": "lin_attn.z", "floor_bits": 4, "assigned_bits": 3, "num_params": 31457280 }, { "name": "model.layers.58.linear_attn.out_proj", "role": "lin_attn.out", "floor_bits": 4, "assigned_bits": 3, "num_params": 31457280 }, { "name": "model.layers.59.mlp.gate_proj", "role": "mlp.gate", "floor_bits": 4, "assigned_bits": 3, "num_params": 89128960 } ], "bits": { "lm_head": 8, "model.embed_tokens": 4, "model.layers.0.linear_attn.in_proj_a": 8, "model.layers.0.linear_attn.in_proj_b": 8, "model.layers.0.linear_attn.in_proj_qkv": 4, "model.layers.0.linear_attn.in_proj_z": 4, "model.layers.0.linear_attn.out_proj": 4, "model.layers.0.mlp.down_proj": 3, "model.layers.0.mlp.gate_proj": 3, "model.layers.0.mlp.up_proj": 2, "model.layers.1.linear_attn.in_proj_a": 8, "model.layers.1.linear_attn.in_proj_b": 8, "model.layers.1.linear_attn.in_proj_qkv": 4, "model.layers.1.linear_attn.in_proj_z": 4, "model.layers.1.linear_attn.out_proj": 4, "model.layers.1.mlp.down_proj": 3, "model.layers.1.mlp.gate_proj": 4, "model.layers.1.mlp.up_proj": 3, "model.layers.10.linear_attn.in_proj_a": 8, "model.layers.10.linear_attn.in_proj_b": 8, "model.layers.10.linear_attn.in_proj_qkv": 4, "model.layers.10.linear_attn.in_proj_z": 4, "model.layers.10.linear_attn.out_proj": 4, "model.layers.10.mlp.down_proj": 3, "model.layers.10.mlp.gate_proj": 4, "model.layers.10.mlp.up_proj": 3, "model.layers.11.mlp.down_proj": 3, "model.layers.11.mlp.gate_proj": 4, "model.layers.11.mlp.up_proj": 3, "model.layers.11.self_attn.k_proj": 4, "model.layers.11.self_attn.o_proj": 4, "model.layers.11.self_attn.q_proj": 4, "model.layers.11.self_attn.v_proj": 4, "model.layers.12.linear_attn.in_proj_a": 8, "model.layers.12.linear_attn.in_proj_b": 8, "model.layers.12.linear_attn.in_proj_qkv": 4, "model.layers.12.linear_attn.in_proj_z": 4, "model.layers.12.linear_attn.out_proj": 4, "model.layers.12.mlp.down_proj": 3, "model.layers.12.mlp.gate_proj": 4, "model.layers.12.mlp.up_proj": 3, "model.layers.13.linear_attn.in_proj_a": 8, "model.layers.13.linear_attn.in_proj_b": 8, "model.layers.13.linear_attn.in_proj_qkv": 4, "model.layers.13.linear_attn.in_proj_z": 4, "model.layers.13.linear_attn.out_proj": 4, "model.layers.13.mlp.down_proj": 3, "model.layers.13.mlp.gate_proj": 4, "model.layers.13.mlp.up_proj": 3, "model.layers.14.linear_attn.in_proj_a": 8, "model.layers.14.linear_attn.in_proj_b": 8, "model.layers.14.linear_attn.in_proj_qkv": 4, "model.layers.14.linear_attn.in_proj_z": 4, "model.layers.14.linear_attn.out_proj": 4, "model.layers.14.mlp.down_proj": 3, "model.layers.14.mlp.gate_proj": 4, "model.layers.14.mlp.up_proj": 3, "model.layers.15.mlp.down_proj": 3, "model.layers.15.mlp.gate_proj": 4, "model.layers.15.mlp.up_proj": 3, "model.layers.15.self_attn.k_proj": 4, "model.layers.15.self_attn.o_proj": 4, "model.layers.15.self_attn.q_proj": 4, "model.layers.15.self_attn.v_proj": 4, "model.layers.16.linear_attn.in_proj_a": 8, "model.layers.16.linear_attn.in_proj_b": 8, "model.layers.16.linear_attn.in_proj_qkv": 4, "model.layers.16.linear_attn.in_proj_z": 4, "model.layers.16.linear_attn.out_proj": 4, "model.layers.16.mlp.down_proj": 3, "model.layers.16.mlp.gate_proj": 4, "model.layers.16.mlp.up_proj": 3, "model.layers.17.linear_attn.in_proj_a": 8, "model.layers.17.linear_attn.in_proj_b": 8, "model.layers.17.linear_attn.in_proj_qkv": 4, "model.layers.17.linear_attn.in_proj_z": 4, "model.layers.17.linear_attn.out_proj": 4, "model.layers.17.mlp.down_proj": 3, "model.layers.17.mlp.gate_proj": 4, "model.layers.17.mlp.up_proj": 3, "model.layers.18.linear_attn.in_proj_a": 8, "model.layers.18.linear_attn.in_proj_b": 8, "model.layers.18.linear_attn.in_proj_qkv": 4, "model.layers.18.linear_attn.in_proj_z": 4, "model.layers.18.linear_attn.out_proj": 4, "model.layers.18.mlp.down_proj": 3, "model.layers.18.mlp.gate_proj": 4, "model.layers.18.mlp.up_proj": 3, "model.layers.19.mlp.down_proj": 3, "model.layers.19.mlp.gate_proj": 4, "model.layers.19.mlp.up_proj": 3, "model.layers.19.self_attn.k_proj": 4, "model.layers.19.self_attn.o_proj": 4, "model.layers.19.self_attn.q_proj": 4, "model.layers.19.self_attn.v_proj": 4, "model.layers.2.linear_attn.in_proj_a": 8, "model.layers.2.linear_attn.in_proj_b": 8, "model.layers.2.linear_attn.in_proj_qkv": 4, "model.layers.2.linear_attn.in_proj_z": 4, "model.layers.2.linear_attn.out_proj": 4, "model.layers.2.mlp.down_proj": 3, "model.layers.2.mlp.gate_proj": 4, "model.layers.2.mlp.up_proj": 3, "model.layers.20.linear_attn.in_proj_a": 8, "model.layers.20.linear_attn.in_proj_b": 8, "model.layers.20.linear_attn.in_proj_qkv": 4, "model.layers.20.linear_attn.in_proj_z": 4, "model.layers.20.linear_attn.out_proj": 4, "model.layers.20.mlp.down_proj": 3, "model.layers.20.mlp.gate_proj": 4, "model.layers.20.mlp.up_proj": 3, "model.layers.21.linear_attn.in_proj_a": 8, "model.layers.21.linear_attn.in_proj_b": 8, "model.layers.21.linear_attn.in_proj_qkv": 4, "model.layers.21.linear_attn.in_proj_z": 4, "model.layers.21.linear_attn.out_proj": 4, "model.layers.21.mlp.down_proj": 3, "model.layers.21.mlp.gate_proj": 4, "model.layers.21.mlp.up_proj": 3, "model.layers.22.linear_attn.in_proj_a": 8, "model.layers.22.linear_attn.in_proj_b": 8, "model.layers.22.linear_attn.in_proj_qkv": 4, "model.layers.22.linear_attn.in_proj_z": 4, "model.layers.22.linear_attn.out_proj": 4, "model.layers.22.mlp.down_proj": 3, "model.layers.22.mlp.gate_proj": 4, "model.layers.22.mlp.up_proj": 3, "model.layers.23.mlp.down_proj": 3, "model.layers.23.mlp.gate_proj": 4, "model.layers.23.mlp.up_proj": 3, "model.layers.23.self_attn.k_proj": 4, "model.layers.23.self_attn.o_proj": 4, "model.layers.23.self_attn.q_proj": 4, "model.layers.23.self_attn.v_proj": 4, "model.layers.24.linear_attn.in_proj_a": 8, "model.layers.24.linear_attn.in_proj_b": 8, "model.layers.24.linear_attn.in_proj_qkv": 4, "model.layers.24.linear_attn.in_proj_z": 4, "model.layers.24.linear_attn.out_proj": 4, "model.layers.24.mlp.down_proj": 3, "model.layers.24.mlp.gate_proj": 4, "model.layers.24.mlp.up_proj": 3, "model.layers.25.linear_attn.in_proj_a": 8, "model.layers.25.linear_attn.in_proj_b": 8, "model.layers.25.linear_attn.in_proj_qkv": 4, "model.layers.25.linear_attn.in_proj_z": 4, "model.layers.25.linear_attn.out_proj": 4, "model.layers.25.mlp.down_proj": 3, "model.layers.25.mlp.gate_proj": 4, "model.layers.25.mlp.up_proj": 3, "model.layers.26.linear_attn.in_proj_a": 8, "model.layers.26.linear_attn.in_proj_b": 8, "model.layers.26.linear_attn.in_proj_qkv": 4, "model.layers.26.linear_attn.in_proj_z": 4, "model.layers.26.linear_attn.out_proj": 4, "model.layers.26.mlp.down_proj": 3, "model.layers.26.mlp.gate_proj": 4, "model.layers.26.mlp.up_proj": 3, "model.layers.27.mlp.down_proj": 3, "model.layers.27.mlp.gate_proj": 4, "model.layers.27.mlp.up_proj": 3, "model.layers.27.self_attn.k_proj": 4, "model.layers.27.self_attn.o_proj": 4, "model.layers.27.self_attn.q_proj": 4, "model.layers.27.self_attn.v_proj": 4, "model.layers.28.linear_attn.in_proj_a": 8, "model.layers.28.linear_attn.in_proj_b": 8, "model.layers.28.linear_attn.in_proj_qkv": 4, "model.layers.28.linear_attn.in_proj_z": 4, "model.layers.28.linear_attn.out_proj": 4, "model.layers.28.mlp.down_proj": 3, "model.layers.28.mlp.gate_proj": 4, "model.layers.28.mlp.up_proj": 3, "model.layers.29.linear_attn.in_proj_a": 8, "model.layers.29.linear_attn.in_proj_b": 8, "model.layers.29.linear_attn.in_proj_qkv": 4, "model.layers.29.linear_attn.in_proj_z": 4, "model.layers.29.linear_attn.out_proj": 4, "model.layers.29.mlp.down_proj": 3, "model.layers.29.mlp.gate_proj": 4, "model.layers.29.mlp.up_proj": 3, "model.layers.3.mlp.down_proj": 3, "model.layers.3.mlp.gate_proj": 4, "model.layers.3.mlp.up_proj": 3, "model.layers.3.self_attn.k_proj": 4, "model.layers.3.self_attn.o_proj": 4, "model.layers.3.self_attn.q_proj": 4, "model.layers.3.self_attn.v_proj": 4, "model.layers.30.linear_attn.in_proj_a": 8, "model.layers.30.linear_attn.in_proj_b": 8, "model.layers.30.linear_attn.in_proj_qkv": 4, "model.layers.30.linear_attn.in_proj_z": 4, "model.layers.30.linear_attn.out_proj": 4, "model.layers.30.mlp.down_proj": 3, "model.layers.30.mlp.gate_proj": 4, "model.layers.30.mlp.up_proj": 3, "model.layers.31.mlp.down_proj": 3, "model.layers.31.mlp.gate_proj": 4, "model.layers.31.mlp.up_proj": 3, "model.layers.31.self_attn.k_proj": 4, "model.layers.31.self_attn.o_proj": 4, "model.layers.31.self_attn.q_proj": 4, "model.layers.31.self_attn.v_proj": 4, "model.layers.32.linear_attn.in_proj_a": 8, "model.layers.32.linear_attn.in_proj_b": 8, "model.layers.32.linear_attn.in_proj_qkv": 4, "model.layers.32.linear_attn.in_proj_z": 4, "model.layers.32.linear_attn.out_proj": 4, "model.layers.32.mlp.down_proj": 3, "model.layers.32.mlp.gate_proj": 4, "model.layers.32.mlp.up_proj": 3, "model.layers.33.linear_attn.in_proj_a": 8, "model.layers.33.linear_attn.in_proj_b": 8, "model.layers.33.linear_attn.in_proj_qkv": 4, "model.layers.33.linear_attn.in_proj_z": 4, "model.layers.33.linear_attn.out_proj": 4, "model.layers.33.mlp.down_proj": 3, "model.layers.33.mlp.gate_proj": 4, "model.layers.33.mlp.up_proj": 3, "model.layers.34.linear_attn.in_proj_a": 8, "model.layers.34.linear_attn.in_proj_b": 8, "model.layers.34.linear_attn.in_proj_qkv": 4, "model.layers.34.linear_attn.in_proj_z": 4, "model.layers.34.linear_attn.out_proj": 4, "model.layers.34.mlp.down_proj": 3, "model.layers.34.mlp.gate_proj": 4, "model.layers.34.mlp.up_proj": 3, "model.layers.35.mlp.down_proj": 3, "model.layers.35.mlp.gate_proj": 4, "model.layers.35.mlp.up_proj": 3, "model.layers.35.self_attn.k_proj": 4, "model.layers.35.self_attn.o_proj": 4, "model.layers.35.self_attn.q_proj": 4, "model.layers.35.self_attn.v_proj": 4, "model.layers.36.linear_attn.in_proj_a": 8, "model.layers.36.linear_attn.in_proj_b": 8, "model.layers.36.linear_attn.in_proj_qkv": 4, "model.layers.36.linear_attn.in_proj_z": 4, "model.layers.36.linear_attn.out_proj": 4, "model.layers.36.mlp.down_proj": 3, "model.layers.36.mlp.gate_proj": 4, "model.layers.36.mlp.up_proj": 3, "model.layers.37.linear_attn.in_proj_a": 8, "model.layers.37.linear_attn.in_proj_b": 8, "model.layers.37.linear_attn.in_proj_qkv": 4, "model.layers.37.linear_attn.in_proj_z": 4, "model.layers.37.linear_attn.out_proj": 4, "model.layers.37.mlp.down_proj": 3, "model.layers.37.mlp.gate_proj": 4, "model.layers.37.mlp.up_proj": 3, "model.layers.38.linear_attn.in_proj_a": 8, "model.layers.38.linear_attn.in_proj_b": 8, "model.layers.38.linear_attn.in_proj_qkv": 4, "model.layers.38.linear_attn.in_proj_z": 4, "model.layers.38.linear_attn.out_proj": 4, "model.layers.38.mlp.down_proj": 3, "model.layers.38.mlp.gate_proj": 4, "model.layers.38.mlp.up_proj": 3, "model.layers.39.mlp.down_proj": 3, "model.layers.39.mlp.gate_proj": 4, "model.layers.39.mlp.up_proj": 3, "model.layers.39.self_attn.k_proj": 4, "model.layers.39.self_attn.o_proj": 4, "model.layers.39.self_attn.q_proj": 4, "model.layers.39.self_attn.v_proj": 4, "model.layers.4.linear_attn.in_proj_a": 8, "model.layers.4.linear_attn.in_proj_b": 8, "model.layers.4.linear_attn.in_proj_qkv": 4, "model.layers.4.linear_attn.in_proj_z": 4, "model.layers.4.linear_attn.out_proj": 4, "model.layers.4.mlp.down_proj": 3, "model.layers.4.mlp.gate_proj": 4, "model.layers.4.mlp.up_proj": 3, "model.layers.40.linear_attn.in_proj_a": 8, "model.layers.40.linear_attn.in_proj_b": 8, "model.layers.40.linear_attn.in_proj_qkv": 4, "model.layers.40.linear_attn.in_proj_z": 4, "model.layers.40.linear_attn.out_proj": 4, "model.layers.40.mlp.down_proj": 3, "model.layers.40.mlp.gate_proj": 4, "model.layers.40.mlp.up_proj": 3, "model.layers.41.linear_attn.in_proj_a": 8, "model.layers.41.linear_attn.in_proj_b": 8, "model.layers.41.linear_attn.in_proj_qkv": 4, "model.layers.41.linear_attn.in_proj_z": 4, "model.layers.41.linear_attn.out_proj": 4, "model.layers.41.mlp.down_proj": 3, "model.layers.41.mlp.gate_proj": 4, "model.layers.41.mlp.up_proj": 3, "model.layers.42.linear_attn.in_proj_a": 8, "model.layers.42.linear_attn.in_proj_b": 8, "model.layers.42.linear_attn.in_proj_qkv": 4, "model.layers.42.linear_attn.in_proj_z": 4, "model.layers.42.linear_attn.out_proj": 4, "model.layers.42.mlp.down_proj": 3, "model.layers.42.mlp.gate_proj": 4, "model.layers.42.mlp.up_proj": 3, "model.layers.43.mlp.down_proj": 3, "model.layers.43.mlp.gate_proj": 4, "model.layers.43.mlp.up_proj": 3, "model.layers.43.self_attn.k_proj": 4, "model.layers.43.self_attn.o_proj": 4, "model.layers.43.self_attn.q_proj": 4, "model.layers.43.self_attn.v_proj": 4, "model.layers.44.linear_attn.in_proj_a": 8, "model.layers.44.linear_attn.in_proj_b": 8, "model.layers.44.linear_attn.in_proj_qkv": 4, "model.layers.44.linear_attn.in_proj_z": 4, "model.layers.44.linear_attn.out_proj": 4, "model.layers.44.mlp.down_proj": 3, "model.layers.44.mlp.gate_proj": 4, "model.layers.44.mlp.up_proj": 3, "model.layers.45.linear_attn.in_proj_a": 8, "model.layers.45.linear_attn.in_proj_b": 8, "model.layers.45.linear_attn.in_proj_qkv": 4, "model.layers.45.linear_attn.in_proj_z": 4, "model.layers.45.linear_attn.out_proj": 4, "model.layers.45.mlp.down_proj": 3, "model.layers.45.mlp.gate_proj": 4, "model.layers.45.mlp.up_proj": 3, "model.layers.46.linear_attn.in_proj_a": 8, "model.layers.46.linear_attn.in_proj_b": 8, "model.layers.46.linear_attn.in_proj_qkv": 4, "model.layers.46.linear_attn.in_proj_z": 4, "model.layers.46.linear_attn.out_proj": 4, "model.layers.46.mlp.down_proj": 3, "model.layers.46.mlp.gate_proj": 4, "model.layers.46.mlp.up_proj": 3, "model.layers.47.mlp.down_proj": 3, "model.layers.47.mlp.gate_proj": 4, "model.layers.47.mlp.up_proj": 3, "model.layers.47.self_attn.k_proj": 4, "model.layers.47.self_attn.o_proj": 4, "model.layers.47.self_attn.q_proj": 4, "model.layers.47.self_attn.v_proj": 4, "model.layers.48.linear_attn.in_proj_a": 8, "model.layers.48.linear_attn.in_proj_b": 8, "model.layers.48.linear_attn.in_proj_qkv": 4, "model.layers.48.linear_attn.in_proj_z": 4, "model.layers.48.linear_attn.out_proj": 4, "model.layers.48.mlp.down_proj": 3, "model.layers.48.mlp.gate_proj": 4, "model.layers.48.mlp.up_proj": 3, "model.layers.49.linear_attn.in_proj_a": 8, "model.layers.49.linear_attn.in_proj_b": 8, "model.layers.49.linear_attn.in_proj_qkv": 4, "model.layers.49.linear_attn.in_proj_z": 4, "model.layers.49.linear_attn.out_proj": 4, "model.layers.49.mlp.down_proj": 3, "model.layers.49.mlp.gate_proj": 4, "model.layers.49.mlp.up_proj": 3, "model.layers.5.linear_attn.in_proj_a": 8, "model.layers.5.linear_attn.in_proj_b": 8, "model.layers.5.linear_attn.in_proj_qkv": 4, "model.layers.5.linear_attn.in_proj_z": 4, "model.layers.5.linear_attn.out_proj": 4, "model.layers.5.mlp.down_proj": 3, "model.layers.5.mlp.gate_proj": 4, "model.layers.5.mlp.up_proj": 3, "model.layers.50.linear_attn.in_proj_a": 8, "model.layers.50.linear_attn.in_proj_b": 8, "model.layers.50.linear_attn.in_proj_qkv": 4, "model.layers.50.linear_attn.in_proj_z": 4, "model.layers.50.linear_attn.out_proj": 4, "model.layers.50.mlp.down_proj": 3, "model.layers.50.mlp.gate_proj": 4, "model.layers.50.mlp.up_proj": 3, "model.layers.51.mlp.down_proj": 3, "model.layers.51.mlp.gate_proj": 4, "model.layers.51.mlp.up_proj": 3, "model.layers.51.self_attn.k_proj": 4, "model.layers.51.self_attn.o_proj": 4, "model.layers.51.self_attn.q_proj": 4, "model.layers.51.self_attn.v_proj": 4, "model.layers.52.linear_attn.in_proj_a": 8, "model.layers.52.linear_attn.in_proj_b": 8, "model.layers.52.linear_attn.in_proj_qkv": 4, "model.layers.52.linear_attn.in_proj_z": 4, "model.layers.52.linear_attn.out_proj": 4, "model.layers.52.mlp.down_proj": 3, "model.layers.52.mlp.gate_proj": 4, "model.layers.52.mlp.up_proj": 3, "model.layers.53.linear_attn.in_proj_a": 8, "model.layers.53.linear_attn.in_proj_b": 8, "model.layers.53.linear_attn.in_proj_qkv": 4, "model.layers.53.linear_attn.in_proj_z": 4, "model.layers.53.linear_attn.out_proj": 4, "model.layers.53.mlp.down_proj": 3, "model.layers.53.mlp.gate_proj": 4, "model.layers.53.mlp.up_proj": 3, "model.layers.54.linear_attn.in_proj_a": 8, "model.layers.54.linear_attn.in_proj_b": 8, "model.layers.54.linear_attn.in_proj_qkv": 3, "model.layers.54.linear_attn.in_proj_z": 4, "model.layers.54.linear_attn.out_proj": 4, "model.layers.54.mlp.down_proj": 3, "model.layers.54.mlp.gate_proj": 4, "model.layers.54.mlp.up_proj": 3, "model.layers.55.mlp.down_proj": 3, "model.layers.55.mlp.gate_proj": 4, "model.layers.55.mlp.up_proj": 3, "model.layers.55.self_attn.k_proj": 4, "model.layers.55.self_attn.o_proj": 3, "model.layers.55.self_attn.q_proj": 3, "model.layers.55.self_attn.v_proj": 4, "model.layers.56.linear_attn.in_proj_a": 8, "model.layers.56.linear_attn.in_proj_b": 8, "model.layers.56.linear_attn.in_proj_qkv": 4, "model.layers.56.linear_attn.in_proj_z": 4, "model.layers.56.linear_attn.out_proj": 4, "model.layers.56.mlp.down_proj": 3, "model.layers.56.mlp.gate_proj": 4, "model.layers.56.mlp.up_proj": 3, "model.layers.57.linear_attn.in_proj_a": 8, "model.layers.57.linear_attn.in_proj_b": 8, "model.layers.57.linear_attn.in_proj_qkv": 4, "model.layers.57.linear_attn.in_proj_z": 4, "model.layers.57.linear_attn.out_proj": 4, "model.layers.57.mlp.down_proj": 3, "model.layers.57.mlp.gate_proj": 4, "model.layers.57.mlp.up_proj": 3, "model.layers.58.linear_attn.in_proj_a": 8, "model.layers.58.linear_attn.in_proj_b": 8, "model.layers.58.linear_attn.in_proj_qkv": 3, "model.layers.58.linear_attn.in_proj_z": 3, "model.layers.58.linear_attn.out_proj": 3, "model.layers.58.mlp.down_proj": 3, "model.layers.58.mlp.gate_proj": 4, "model.layers.58.mlp.up_proj": 3, "model.layers.59.mlp.down_proj": 3, "model.layers.59.mlp.gate_proj": 3, "model.layers.59.mlp.up_proj": 3, "model.layers.59.self_attn.k_proj": 4, "model.layers.59.self_attn.o_proj": 4, "model.layers.59.self_attn.q_proj": 4, "model.layers.59.self_attn.v_proj": 4, "model.layers.6.linear_attn.in_proj_a": 8, "model.layers.6.linear_attn.in_proj_b": 8, "model.layers.6.linear_attn.in_proj_qkv": 4, "model.layers.6.linear_attn.in_proj_z": 4, "model.layers.6.linear_attn.out_proj": 4, "model.layers.6.mlp.down_proj": 3, "model.layers.6.mlp.gate_proj": 4, "model.layers.6.mlp.up_proj": 3, "model.layers.60.linear_attn.in_proj_a": 8, "model.layers.60.linear_attn.in_proj_b": 8, "model.layers.60.linear_attn.in_proj_qkv": 4, "model.layers.60.linear_attn.in_proj_z": 4, "model.layers.60.linear_attn.out_proj": 4, "model.layers.60.mlp.down_proj": 3, "model.layers.60.mlp.gate_proj": 4, "model.layers.60.mlp.up_proj": 3, "model.layers.61.linear_attn.in_proj_a": 8, "model.layers.61.linear_attn.in_proj_b": 8, "model.layers.61.linear_attn.in_proj_qkv": 4, "model.layers.61.linear_attn.in_proj_z": 4, "model.layers.61.linear_attn.out_proj": 4, "model.layers.61.mlp.down_proj": 3, "model.layers.61.mlp.gate_proj": 4, "model.layers.61.mlp.up_proj": 3, "model.layers.62.linear_attn.in_proj_a": 8, "model.layers.62.linear_attn.in_proj_b": 8, "model.layers.62.linear_attn.in_proj_qkv": 4, "model.layers.62.linear_attn.in_proj_z": 4, "model.layers.62.linear_attn.out_proj": 4, "model.layers.62.mlp.down_proj": 3, "model.layers.62.mlp.gate_proj": 4, "model.layers.62.mlp.up_proj": 3, "model.layers.63.mlp.down_proj": 3, "model.layers.63.mlp.gate_proj": 4, "model.layers.63.mlp.up_proj": 3, "model.layers.63.self_attn.k_proj": 4, "model.layers.63.self_attn.o_proj": 4, "model.layers.63.self_attn.q_proj": 4, "model.layers.63.self_attn.v_proj": 4, "model.layers.7.mlp.down_proj": 3, "model.layers.7.mlp.gate_proj": 4, "model.layers.7.mlp.up_proj": 3, "model.layers.7.self_attn.k_proj": 4, "model.layers.7.self_attn.o_proj": 4, "model.layers.7.self_attn.q_proj": 4, "model.layers.7.self_attn.v_proj": 4, "model.layers.8.linear_attn.in_proj_a": 8, "model.layers.8.linear_attn.in_proj_b": 8, "model.layers.8.linear_attn.in_proj_qkv": 4, "model.layers.8.linear_attn.in_proj_z": 4, "model.layers.8.linear_attn.out_proj": 4, "model.layers.8.mlp.down_proj": 3, "model.layers.8.mlp.gate_proj": 4, "model.layers.8.mlp.up_proj": 3, "model.layers.9.linear_attn.in_proj_a": 8, "model.layers.9.linear_attn.in_proj_b": 8, "model.layers.9.linear_attn.in_proj_qkv": 4, "model.layers.9.linear_attn.in_proj_z": 4, "model.layers.9.linear_attn.out_proj": 4, "model.layers.9.mlp.down_proj": 3, "model.layers.9.mlp.gate_proj": 4, "model.layers.9.mlp.up_proj": 3 } } } }