Llama-3.1-8B-Instruct-AutoRound-Recipe / llama3.1-8b-mxfp4-mixed-config.json
xinhe's picture
Upload llama3.1-8b-mxfp4-mixed-config.json
ef2d408 verified
Raw
History Blame Contribute Delete
20.4 kB
{
"architectures": [
"LlamaForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 128000,
"dtype": "bfloat16",
"eos_token_id": [
128001,
128008,
128009
],
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 4096,
"initializer_range": 0.02,
"intermediate_size": 14336,
"max_position_embeddings": 131072,
"mlp_bias": false,
"model_type": "llama",
"num_attention_heads": 32,
"num_hidden_layers": 32,
"num_key_value_heads": 8,
"pretraining_tp": 1,
"quantization_config": {
"act_bits": 4,
"act_data_type": "mx_fp_rceil",
"act_dynamic": true,
"act_group_size": 32,
"act_sym": true,
"autoround_version": "0.12.0",
"bits": 4,
"data_type": "mx_fp",
"extra_config": {
"model.layers.0.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.0.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.1.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.10.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.11.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.12.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.13.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.14.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.15.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.16.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.17.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.18.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.19.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.2.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.20.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.21.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.22.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.23.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.24.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.24.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.24.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.24.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.25.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.25.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.25.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.25.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.26.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.26.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.26.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.26.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.26.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.27.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.27.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.27.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.27.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.28.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.28.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.28.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.28.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.28.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.29.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.29.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.29.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.29.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.29.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.3.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.30.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.31.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.4.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.5.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.6.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.7.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.8.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.mlp.down_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.mlp.gate_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.mlp.up_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.self_attn.k_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.self_attn.o_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.self_attn.q_proj": {
"act_bits": 8,
"bits": 8
},
"model.layers.9.self_attn.v_proj": {
"act_bits": 8,
"bits": 8
}
},
"group_size": 32,
"iters": 0,
"low_gpu_mem_usage": true,
"packing_format": "auto_round:llm_compressor",
"quant_method": "auto-round",
"sym": true
},
"rms_norm_eps": 1e-05,
"rope_scaling": {
"factor": 8.0,
"high_freq_factor": 4.0,
"low_freq_factor": 1.0,
"original_max_position_embeddings": 8192,
"rope_type": "llama3"
},
"rope_theta": 500000.0,
"tie_word_embeddings": false,
"transformers_version": "4.57.3",
"use_cache": true,
"vocab_size": 128256
}