nimble:9b-nvfp4

11.4K 3 hours ago

A 9B decision model from Bespoke Labs for fast, typed classification.

tools decision thinking 9b
... /
json
b25d9fe2dfda · 32kB
{
"producer": {
"name": "modelopt",
"version": "0.47.0.dev75+gfbcdc16c2.d20260818"
},
"quantization": {
"quant_algo": "MIXED_PRECISION",
"kv_cache_quant_algo": null,
"quantized_layers": {
"model.language_model.embed_tokens": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.0.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.0.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.0.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.0.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.0.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.0.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.1.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.1.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.1.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.1.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.1.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.1.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.2.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.2.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.2.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.2.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.2.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.2.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.3.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.3.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.3.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.3.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.3.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.3.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.3.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.4.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.4.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.4.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.4.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.4.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.4.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.5.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.5.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.5.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.5.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.5.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.5.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.6.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.6.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.6.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.6.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.6.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.6.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.7.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.7.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.7.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.7.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.7.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.7.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.7.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.8.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.8.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.8.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.8.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.8.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.8.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.9.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.9.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.9.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.9.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.9.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.9.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.10.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.10.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.10.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.10.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.10.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.10.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.11.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.11.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.11.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.11.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.11.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.11.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.11.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.12.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.12.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.12.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.12.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.12.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.12.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.13.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.13.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.13.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.13.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.13.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.13.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.14.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.14.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.14.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.14.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.14.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.14.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.15.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.15.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.15.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.15.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.15.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.15.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.15.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.16.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.16.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.16.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.16.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.16.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.16.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.17.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.17.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.17.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.17.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.17.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.17.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.18.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.18.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.18.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.18.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.18.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.18.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.19.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.19.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.19.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.19.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.19.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.19.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.19.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.20.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.20.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.20.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.20.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.20.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.20.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.21.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.21.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.21.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.21.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.21.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.21.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.22.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.22.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.22.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.22.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.22.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.22.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.23.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.23.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.23.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.23.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.23.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.23.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.23.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.24.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.24.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.24.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.24.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.24.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.24.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.25.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.25.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.25.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.25.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.25.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.25.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.26.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.26.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.26.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.26.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.26.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.26.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.27.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.27.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.27.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.27.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.27.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.27.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.27.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.28.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.28.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.28.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.28.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.28.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.28.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.29.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.29.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.29.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.29.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.29.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.29.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.30.linear_attn.out_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.30.linear_attn.in_proj_qkv": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.30.linear_attn.in_proj_z": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.30.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.30.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.30.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.31.self_attn.q_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.31.self_attn.k_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.31.self_attn.v_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"model.language_model.layers.31.self_attn.o_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.31.mlp.gate_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.31.mlp.up_proj": {
"quant_algo": "W4A16_NVFP4",
"group_size": 16
},
"model.language_model.layers.31.mlp.down_proj": {
"quant_algo": "MXFP8",
"group_size": 32
},
"lm_head": {
"quant_algo": "MXFP8",
"group_size": 32
}
},
"exclude_modules": [
"mtp*",
"mtp.layers.0*"
]
}
}