Step-3.5-MoE-NVFP4 / hf_quant_config.json
apandacoding's picture
Upload folder using huggingface_hub
0408287 verified
{
"producer": {
"name": "modelopt",
"version": "0.41.0"
},
"quantization": {
"quant_algo": "MIXED_PRECISION",
"kv_cache_quant_algo": null,
"quantized_layers": {
"model.layers.0.mlp.up_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.0.mlp.down_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.1.mlp.up_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.1.mlp.down_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.2.mlp.up_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.2.mlp.down_proj": {
"quant_algo": "W8A8_SQ_PER_CHANNEL"
},
"model.layers.3.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.3.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.4.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.5.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.6.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.7.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.8.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.9.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.10.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.11.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.12.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.13.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.14.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.15.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.16.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.17.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.18.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.19.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.20.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.21.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.22.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.23.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.24.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.25.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.26.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.27.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.28.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.29.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.30.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.31.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.32.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.33.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.34.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.35.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.36.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.37.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.38.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.39.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.40.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.41.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.42.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.43.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.moe.up_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.moe.gate_proj": {
"quant_algo": "NVFP4",
"group_size": 16
},
"model.layers.44.moe.down_proj": {
"quant_algo": "NVFP4",
"group_size": 16
}
}
}
}