| { | |
| "config": { | |
| "vocab_size": 32000, | |
| "max_position_embeddings": 4096, | |
| "hidden_size": 4096, | |
| "intermediate_size": 11008, | |
| "num_hidden_layers": 32, | |
| "num_attention_heads": 32, | |
| "num_key_value_heads": 32, | |
| "hidden_act": "silu", | |
| "initializer_range": 0.02, | |
| "rms_norm_eps": 1e-05, | |
| "pretraining_tp": 1, | |
| "use_cache": true, | |
| "rope_theta": 10000.0, | |
| "rope_scaling": null, | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "torch_dtype": "float16", | |
| "tie_word_embeddings": false, | |
| "architectures": [ | |
| "LlamaForCausalLM" | |
| ], | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "_name_or_path": "meta-llama/Llama-2-7b-chat-hf", | |
| "transformers_version": "4.37.2", | |
| "model_type": "llama" | |
| }, | |
| "structure": { | |
| "model": { | |
| "embed_tokens": { | |
| "weight": "model.embed_tokens.weight", | |
| "type": "Embedding" | |
| }, | |
| "layers": { | |
| "0": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.0.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.0.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.0.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.0.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.0.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.0.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.0.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.0.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.0.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.0.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.0.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.0.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.0.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.0.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.0.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.0.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.0.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "1": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.1.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.1.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.1.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.1.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.1.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.1.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.1.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.1.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.1.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.1.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.1.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.1.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.1.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.1.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.1.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.1.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.1.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "2": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.2.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.2.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.2.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.2.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.2.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.2.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.2.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.2.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.2.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.2.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.2.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.2.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.2.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.2.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.2.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.2.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.2.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "3": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.3.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.3.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.3.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.3.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.3.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.3.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.3.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.3.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.3.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.3.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.3.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.3.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.3.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.3.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.3.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.3.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.3.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "4": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.4.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.4.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.4.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.4.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.4.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.4.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.4.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.4.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.4.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.4.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.4.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.4.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.4.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.4.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.4.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.4.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.4.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "5": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.5.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.5.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.5.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.5.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.5.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.5.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.5.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.5.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.5.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.5.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.5.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.5.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.5.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.5.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.5.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.5.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.5.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "6": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.6.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.6.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.6.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.6.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.6.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.6.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.6.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.6.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.6.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.6.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.6.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.6.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.6.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.6.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.6.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.6.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.6.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "7": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.7.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.7.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.7.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.7.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.7.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.7.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.7.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.7.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.7.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.7.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.7.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.7.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.7.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.7.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.7.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.7.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.7.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "8": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.8.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.8.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.8.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.8.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.8.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.8.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.8.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.8.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.8.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.8.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.8.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.8.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.8.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.8.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.8.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.8.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.8.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "9": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.9.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.9.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.9.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.9.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.9.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.9.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.9.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.9.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.9.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.9.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.9.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.9.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.9.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.9.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.9.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.9.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.9.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "10": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.10.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.10.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.10.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.10.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.10.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.10.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.10.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.10.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.10.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.10.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.10.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.10.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.10.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.10.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.10.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.10.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.10.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "11": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.11.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.11.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.11.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.11.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.11.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.11.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.11.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.11.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.11.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.11.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.11.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.11.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.11.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.11.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.11.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.11.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.11.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "12": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.12.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.12.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.12.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.12.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.12.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.12.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.12.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.12.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.12.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.12.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.12.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.12.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.12.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.12.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.12.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.12.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.12.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "13": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.13.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.13.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.13.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.13.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.13.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.13.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.13.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.13.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.13.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.13.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.13.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.13.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.13.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.13.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.13.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.13.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.13.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "14": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.14.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.14.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.14.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.14.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.14.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.14.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.14.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.14.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.14.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.14.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.14.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.14.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.14.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.14.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.14.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.14.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.14.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "15": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.15.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.15.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.15.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.15.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.15.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.15.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.15.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.15.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.15.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.15.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.15.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.15.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.15.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.15.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.15.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.15.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.15.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "16": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.16.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.16.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.16.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.16.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.16.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.16.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.16.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.16.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.16.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.16.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.16.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.16.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.16.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.16.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.16.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.16.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.16.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "17": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.17.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.17.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.17.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.17.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.17.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.17.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.17.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.17.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.17.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.17.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.17.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.17.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.17.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.17.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.17.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.17.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.17.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "18": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.18.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.18.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.18.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.18.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.18.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.18.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.18.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.18.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.18.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.18.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.18.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.18.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.18.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.18.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.18.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.18.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.18.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "19": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.19.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.19.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.19.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.19.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.19.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.19.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.19.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.19.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.19.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.19.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.19.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.19.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.19.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.19.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.19.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.19.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.19.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "20": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.20.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.20.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.20.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.20.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.20.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.20.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.20.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.20.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.20.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.20.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.20.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.20.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.20.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.20.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.20.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.20.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.20.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "21": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.21.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.21.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.21.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.21.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.21.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.21.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.21.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.21.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.21.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.21.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.21.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.21.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.21.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.21.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.21.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.21.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.21.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "22": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.22.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.22.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.22.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.22.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.22.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.22.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.22.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.22.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.22.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.22.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.22.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.22.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.22.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.22.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.22.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.22.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.22.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "23": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.23.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.23.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.23.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.23.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.23.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.23.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.23.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.23.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.23.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.23.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.23.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.23.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.23.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.23.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.23.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.23.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.23.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "24": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.24.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.24.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.24.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.24.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.24.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.24.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.24.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.24.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.24.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.24.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.24.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.24.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.24.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.24.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.24.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.24.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.24.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "25": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.25.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.25.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.25.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.25.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.25.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.25.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.25.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.25.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.25.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.25.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.25.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.25.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.25.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.25.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.25.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.25.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.25.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "26": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.26.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.26.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.26.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.26.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.26.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.26.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.26.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.26.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.26.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.26.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.26.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.26.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.26.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.26.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.26.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.26.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.26.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "27": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.27.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.27.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.27.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.27.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.27.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.27.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.27.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.27.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.27.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.27.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.27.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.27.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.27.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.27.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.27.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.27.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.27.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "28": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.28.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.28.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.28.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.28.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.28.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.28.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.28.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.28.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.28.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.28.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.28.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.28.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.28.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.28.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.28.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.28.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.28.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "29": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.29.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.29.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.29.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.29.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.29.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.29.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.29.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.29.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.29.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.29.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.29.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.29.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.29.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.29.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.29.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.29.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.29.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "30": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.30.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.30.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.30.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.30.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.30.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.30.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.30.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.30.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.30.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.30.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.30.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.30.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.30.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.30.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.30.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.30.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.30.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "31": { | |
| "self_attn": { | |
| "q_proj": { | |
| "weight": "model.layers.31.self_attn.q_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.self_attn.q_proj.weight_scale", | |
| "zero_point": "model.layers.31.self_attn.q_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "k_proj": { | |
| "weight": "model.layers.31.self_attn.k_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.self_attn.k_proj.weight_scale", | |
| "zero_point": "model.layers.31.self_attn.k_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "v_proj": { | |
| "weight": "model.layers.31.self_attn.v_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.self_attn.v_proj.weight_scale", | |
| "zero_point": "model.layers.31.self_attn.v_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "o_proj": { | |
| "weight": "model.layers.31.self_attn.o_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.self_attn.o_proj.weight_scale", | |
| "zero_point": "model.layers.31.self_attn.o_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "rotary_emb": { | |
| "type": "LlamaRotaryEmbedding" | |
| } | |
| }, | |
| "mlp": { | |
| "gate_proj": { | |
| "weight": "model.layers.31.mlp.gate_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.mlp.gate_proj.weight_scale", | |
| "zero_point": "model.layers.31.mlp.gate_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "up_proj": { | |
| "weight": "model.layers.31.mlp.up_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.mlp.up_proj.weight_scale", | |
| "zero_point": "model.layers.31.mlp.up_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "down_proj": { | |
| "weight": "model.layers.31.mlp.down_proj.weight", | |
| "weight_quant": { | |
| "scale": "model.layers.31.mlp.down_proj.weight_scale", | |
| "zero_point": "model.layers.31.mlp.down_proj.weight_zero_point", | |
| "dtype": "int4", | |
| "qscheme": "per_group", | |
| "ch_axis": 0, | |
| "group_size": 128, | |
| "round_method": "half_even", | |
| "scale_type": "float" | |
| }, | |
| "type": "QuantLinear" | |
| }, | |
| "act_fn": { | |
| "type": "SiLU" | |
| } | |
| }, | |
| "input_layernorm": { | |
| "weight": "model.layers.31.input_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| }, | |
| "post_attention_layernorm": { | |
| "weight": "model.layers.31.post_attention_layernorm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| } | |
| }, | |
| "norm": { | |
| "weight": "model.norm.weight", | |
| "type": "LlamaRMSNorm" | |
| } | |
| }, | |
| "lm_head": { | |
| "weight": "lm_head.weight", | |
| "type": "Linear" | |
| } | |
| } | |
| } |