151 lines
6.6 KiB
CSV
151 lines
6.6 KiB
CSV
/home/breadway/.cache/breadsearch/models/model.onnx quantization quantization info,2026-06-24 18:14:20
|
|
quantization stage,time consumed(s),sub stage,time consumed(s)
|
|
pre process,5.43
|
|
,,calibration: collect data (onnx inference + numpy statistics),16.60
|
|
,,calibration: compute data,0.01
|
|
calibration (collect data + compute data),19.75
|
|
static quantization,0.81
|
|
post process(including finetuning),0.00
|
|
e2e,40.621040006999465
|
|
|
|
Node Name,Op Type,Activation,Weights,Bias
|
|
/encoder/layers.0/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.0/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.0/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.0/Add,Add,INT8,,
|
|
/encoder/layers.0/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.0/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.0/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.0/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.0/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.0/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.0/Add_1,Add,INT8,,
|
|
/encoder/layers.1/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.1/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.1/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.1/Add,Add,INT8,,
|
|
/encoder/layers.1/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.1/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.1/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.1/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.1/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.1/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.1/Add_1,Add,INT8,,
|
|
/encoder/layers.2/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.2/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.2/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.2/Add,Add,INT8,,
|
|
/encoder/layers.2/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.2/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.2/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.2/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.2/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.2/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.2/Add_1,Add,INT8,,
|
|
/encoder/layers.3/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.3/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.3/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.3/Add,Add,INT8,,
|
|
/encoder/layers.3/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.3/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.3/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.3/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.3/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.3/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.3/Add_1,Add,INT8,,
|
|
/encoder/layers.4/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.4/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.4/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.4/Add,Add,INT8,,
|
|
/encoder/layers.4/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.4/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.4/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.4/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.4/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.4/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.4/Add_1,Add,INT8,,
|
|
/encoder/layers.5/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.5/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.5/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.5/Add,Add,INT8,,
|
|
/encoder/layers.5/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.5/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.5/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.5/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.5/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.5/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.5/Add_1,Add,INT8,,
|
|
/encoder/layers.6/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.6/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.6/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.6/Add,Add,INT8,,
|
|
/encoder/layers.6/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.6/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.6/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.6/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.6/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.6/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.6/Add_1,Add,INT8,,
|
|
/encoder/layers.7/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.7/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.7/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.7/Add,Add,INT8,,
|
|
/encoder/layers.7/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.7/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.7/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.7/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.7/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.7/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.7/Add_1,Add,INT8,,
|
|
/encoder/layers.8/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.8/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.8/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.8/Add,Add,INT8,,
|
|
/encoder/layers.8/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.8/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.8/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.8/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.8/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.8/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.8/Add_1,Add,INT8,,
|
|
/encoder/layers.9/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.9/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.9/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.9/Add,Add,INT8,,
|
|
/encoder/layers.9/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.9/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.9/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.9/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.9/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.9/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.9/Add_1,Add,INT8,,
|
|
/encoder/layers.10/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.10/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.10/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.10/Add,Add,INT8,,
|
|
/encoder/layers.10/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.10/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.10/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.10/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.10/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.10/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.10/Add_1,Add,INT8,,
|
|
/encoder/layers.11/attn/Wqkv/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.11/attn/Reshape,Reshape,INT8,,
|
|
/encoder/layers.11/attn/out_proj/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.11/Add,Add,INT8,,
|
|
/encoder/layers.11/mlp/fc12/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.11/mlp/fc11/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.11/mlp/Sigmoid,Sigmoid,INT8,,
|
|
/encoder/layers.11/mlp/Mul,Mul,INT8,,
|
|
/encoder/layers.11/mlp/Mul_1,Mul,INT8,,
|
|
/encoder/layers.11/mlp/fc2/MatMul,MatMul,INT8,INT8,
|
|
/encoder/layers.11/Add_1,Add,INT8,,
|
|
|
|
Op Type,Activation,Weights,Bias
|
|
MatMul,INT8(60),INT8(60),
|
|
Reshape,INT8(12),,
|
|
Add,INT8(24),,
|
|
Sigmoid,INT8(12),,
|
|
Mul,INT8(24),,
|
|
|