DeepSeek-V2-Lite-int4-AutoRound / quantization_config.json
weiweiz1's picture
Upload folder using huggingface_hub
6e2e49a verified
Raw
History Blame Contribute Delete
10.9 kB
{
"bits": 4,
"data_type": "int",
"group_size": 128,
"sym": true,
"autoround_version": "0.14.0",
"block_name_to_quantize": "model.layers",
"quant_method": "auto-round",
"packing_format": "auto_round:auto_gptq",
"extra_config": {
"model.layers.0.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.0.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.0.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.0.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.1.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.1.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.1.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.1.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.2.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.2.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.2.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.2.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.3.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.3.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.3.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.3.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.4.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.4.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.4.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.4.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.5.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.5.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.5.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.5.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.6.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.6.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.6.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.6.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.7.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.7.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.7.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.7.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.8.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.8.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.8.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.8.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.9.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.9.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.9.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.9.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.10.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.10.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.10.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.10.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.11.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.11.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.11.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.11.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.12.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.12.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.12.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.12.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.13.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.13.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.13.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.13.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.14.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.14.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.14.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.14.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.15.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.15.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.15.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.15.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.16.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.16.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.16.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.16.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.17.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.17.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.17.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.17.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.18.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.18.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.18.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.18.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.19.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.19.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.19.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.19.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.20.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.20.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.20.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.20.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.21.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.21.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.21.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.21.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.22.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.22.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.22.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.22.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.23.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.23.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.23.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.23.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.24.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.24.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.24.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.24.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.25.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.25.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.25.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.25.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.26.self_attn.q_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.26.self_attn.kv_a_proj_with_mqa": {
"bits": 16,
"data_type": "float"
},
"model.layers.26.self_attn.kv_b_proj": {
"bits": 16,
"data_type": "float"
},
"model.layers.26.self_attn.o_proj": {
"bits": 16,
"data_type": "float"
},
".*self_attn.*": {
"bits": 16,
"data_type": "float"
},
".*model\\.layers\\.[1-26]\\.mlp\\.gate.*": {
"bits": 16,
"data_type": "float"
}
}
}