| { | |
| "version": 1.2, | |
| "quantization": "int8_sq", | |
| "dtype": "float16", | |
| "vocab_size": 32000, | |
| "rank": 0, | |
| "tensor_parallel": 1, | |
| "pipeline_parallel": 1, | |
| "vocab_embedding": { | |
| "weight": "transformer.vocab_embedding.weight" | |
| }, | |
| "positional_embedding": null, | |
| "ln_embed": null, | |
| "layers": [ | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.0.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.0.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.0.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.0.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.0.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.0.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.0.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.0.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.1.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.1.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.1.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.1.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.1.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.1.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.1.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.1.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.2.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.2.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.2.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.2.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.2.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.2.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.2.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.2.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.3.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.3.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.3.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.3.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.3.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.3.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.3.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.3.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.4.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.4.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.4.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.4.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.4.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.4.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.4.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.4.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.5.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.5.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.5.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.5.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.5.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.5.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.5.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.5.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.6.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.6.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.6.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.6.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.6.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.6.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.6.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.6.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.7.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.7.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.7.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.7.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.7.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.7.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.7.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.7.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.8.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.8.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.8.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.8.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.8.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.8.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.8.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.8.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.9.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.9.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.9.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.9.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.9.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.9.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.9.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.9.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.10.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.10.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.10.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.10.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.10.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.10.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.10.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.10.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.11.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.11.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.11.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.11.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.11.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.11.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.11.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.11.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.12.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.12.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.12.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.12.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.12.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.12.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.12.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.12.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.13.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.13.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.13.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.13.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.13.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.13.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.13.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.13.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.14.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.14.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.14.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.14.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.14.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.14.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.14.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.14.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.15.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.15.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.15.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.15.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.15.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.15.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.15.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.15.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.16.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.16.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.16.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.16.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.16.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.16.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.16.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.16.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.17.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.17.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.17.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.17.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.17.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.17.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.17.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.17.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.18.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.18.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.18.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.18.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.18.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.18.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.18.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.18.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.19.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.19.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.19.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.19.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.19.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.19.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.19.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.19.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.20.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.20.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.20.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.20.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.20.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.20.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.20.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.20.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.21.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.21.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.21.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.21.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.21.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.21.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.21.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.21.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.22.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.22.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.22.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.22.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.22.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.22.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.22.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.22.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.23.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.23.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.23.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.23.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.23.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.23.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.23.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.23.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.24.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.24.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.24.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.24.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.24.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.24.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.24.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.24.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.25.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.25.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.25.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.25.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.25.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.25.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.25.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.25.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.26.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.26.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.26.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.26.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.26.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.26.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.26.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.26.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.27.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.27.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.27.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.27.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.27.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.27.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.27.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.27.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.28.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.28.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.28.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.28.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.28.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.28.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.28.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.28.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.29.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.29.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.29.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.29.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.29.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.29.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.29.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.29.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.30.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.30.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.30.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.30.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.30.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.30.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.30.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.30.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| }, | |
| { | |
| "quantization": "int8_sq", | |
| "decoder_type": "llama", | |
| "input_layernorm": { | |
| "weight": "transformer.layers.31.input_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp_layernorm": null, | |
| "attention": { | |
| "qkv": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.31.attention.qkv.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "dense": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.31.attention.dense.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "kv_cache_scaling_factor": "transformer.layers.31.attention.kv_cache_scaling_factor", | |
| "kv_cache_dtype": "FP8", | |
| "rotary_dim": -Infinity, | |
| "clip_qkv": null | |
| }, | |
| "post_layernorm": { | |
| "weight": "transformer.layers.31.post_layernorm.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "mlp": { | |
| "fc": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.31.mlp.fc.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "gate": { | |
| "linear_type": "column", | |
| "weight": "transformer.layers.31.mlp.gate.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "proj": { | |
| "linear_type": "row", | |
| "weight": "transformer.layers.31.mlp.proj.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "hidden_act": "silu" | |
| }, | |
| "num_attention_heads": 32, | |
| "num_kv_heads": 32, | |
| "max_position_embeddings": 4096, | |
| "rotary_pct": 0, | |
| "rotary_emb_base": 10000, | |
| "use_alibi": false, | |
| "new_decoder_architecture": false, | |
| "parallel_attention": false, | |
| "apply_residual_connection_post_layernorm": false, | |
| "use_cache": true, | |
| "model_name": "_", | |
| "rope_ratio": 1.0, | |
| "seq_length": 0 | |
| } | |
| ], | |
| "final_layernorm": null, | |
| "ln_f": { | |
| "weight": "transformer.ln_f.weight", | |
| "bias": null, | |
| "layernorm_type": "rms", | |
| "eps": 1e-05 | |
| }, | |
| "lm_head": { | |
| "linear_type": "column", | |
| "weight": "lm_head.weight", | |
| "bias": null, | |
| "activation_scaling_factor": null, | |
| "weights_scaling_factor": null, | |
| "weights_scaling_factor_2": null, | |
| "prequant_scaling_factor": null, | |
| "awq_block_size": 0 | |
| }, | |
| "share_embedding_table": false | |
| } |