forked from wangye707/inference_benchmark
-
Notifications
You must be signed in to change notification settings - Fork 0
/
result2xlsx.py
218 lines (196 loc) · 8.46 KB
/
result2xlsx.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
# -*- coding: utf-8 -*-
# encoding=utf-8 vi:ts=4:sw=4:expandtab:ft=python
import time
import paddle
import argparse
import pandas as pd
import onnxruntime
def parse_args():
"""
parse input args
"""
parser = argparse.ArgumentParser()
parser.add_argument(
"--result_path",
type=str,
default="./result.txt",
help="tipc benchmark log path")
parser.add_argument(
"--output_name",
type=str,
default="tipc_benchmark_excel.xlsx",
help="output excel file name")
parser.add_argument(
"--docker_name",
type=str,
default="paddle_manylinux_devel:cuda11.2-cudnn8.2.1-trt8.0.3.4-gcc82",
help="docker name")
return parser.parse_args()
def log_split(file_name: str) -> list:
"""
log split
"""
log_list = []
with open(file_name, 'r') as f:
log_lines = f.read().split("model path")
for log_line in log_lines:
log_list.append(log_line)
return log_list
def model2onnx_log(file_name: str) -> list:
"""
model to onnx log split
"""
model2onnx_log_list = []
with open(file_name, 'r') as f:
model2onnx_log = f.readlines()
for log_line in model2onnx_log:
if "convert" in log_line:
model2onnx_log_list.append(log_line)
return model2onnx_log_list
def process_log(log_list: list) -> dict:
"""
process log to dict
"""
output_dict = {}
for log_line in log_list.split("\n"):
if "model_name" in log_line:
output_dict["model_name"] = log_line.split(" : ")[-1].strip()
continue
if "cpu_threads" in log_line:
output_dict["cpu_threads"] = log_line.split(" : ")[-1].strip()
continue
if "enable_mkldnn" in log_line:
output_dict["enable_mkldnn"] = log_line.split(" : ")[-1].strip()
continue
if "avg_cost" in log_line:
output_dict["avg_cost"] = log_line.split(" : ")[-1].strip()
continue
if "device_name" in log_line:
output_dict["device_name"] = log_line.split(" : ")[-1].strip()
continue
if "gpu_mem" in log_line:
output_dict["gpu_mem"] = log_line.split(" : ")[-1].strip()
continue
if "backend_type" in log_line:
output_dict["backend_type"] = log_line.split(" : ")[-1].strip()
continue
if "batch_size" in log_line:
output_dict["batch_size"] = log_line.split(" : ")[-1].strip()
continue
if "precision" in log_line:
output_dict["precision"] = log_line.split(" : ")[-1].strip()
continue
if "enable_gpu" in log_line:
output_dict["enable_gpu"] = log_line.split(" : ")[-1].strip()
continue
if "enable_trt" in log_line:
output_dict["enable_trt"] = log_line.split(" : ")[-1].strip()
continue
if "cpu_mem" in log_line:
output_dict["cpu_mem"] = log_line.split(" : ")[-1].strip()
continue
else:
continue
if "device_name" not in output_dict.keys(
) and "model_name" in output_dict.keys():
output_dict["device_name"] = "CPU"
output_dict["gpu_mem"] = "0"
return output_dict
def data_merging(env, paddle_version, onnxruntime_version,
model2onnx_name_list, output_total_list):
log_time = time.strftime('%Y-%m-%d', time.localtime(time.time()))
data_merging_logs = []
for output_dict_num in range(len(output_total_list)):
output_dict = output_total_list[output_dict_num]
log_dict = {}
for comp_list in output_total_list[output_dict_num:]:
if output_dict["model_name"] == comp_list["model_name"] and \
output_dict["device_name"] == comp_list["device_name"] and \
output_dict["batch_size"] == comp_list["batch_size"] and \
output_dict["precision"] == comp_list["precision"] and \
output_dict["cpu_threads"] == comp_list["cpu_threads"] and \
output_dict["enable_mkldnn"] == comp_list["enable_mkldnn"] and \
output_dict["enable_gpu"] == comp_list["enable_gpu"] and \
output_dict["enable_trt"] == comp_list["enable_trt"] and \
output_dict["backend_type"] != comp_list["backend_type"]:
log_dict["日期"] = log_time
log_dict["环境"] = env
log_dict["paddle_version"] = paddle_version
log_dict["onnxruntime_version"] = onnxruntime_version
log_dict["device_name"] = output_dict["device_name"]
log_dict["model_name"] = output_dict["model_name"]
log_dict["precision"] = output_dict["precision"]
log_dict["batch_size"] = output_dict["batch_size"]
log_dict["enable_gpu"] = output_dict["enable_gpu"]
log_dict["enable_mkldnn"] = output_dict["enable_mkldnn"]
log_dict["cpu_threads"] = output_dict["cpu_threads"]
log_dict["enable_trt"] = output_dict["enable_trt"]
log_dict["onnxruntime_cpu_mem"] = output_dict["cpu_mem"]
log_dict["paddle_cpu_mem"] = comp_list["cpu_mem"]
log_dict["paddle_gpu_mem"] = comp_list["gpu_mem"]
log_dict["onnxruntime_gpu_mem"] = output_dict["gpu_mem"]
log_dict["paddle_avg_cost"] = comp_list["avg_cost"]
log_dict["onnxruntime_avg_cost"] = output_dict["avg_cost"]
log_dict["cpu_mem_gap(%)"] = calculation_gap(
paddle_num=log_dict["paddle_cpu_mem"],
onnxruntime_num=log_dict["paddle_cpu_mem"])
log_dict["gpu_mem_gap(%)"] = calculation_gap(
paddle_num=log_dict["paddle_gpu_mem"],
onnxruntime_num=log_dict["onnxruntime_gpu_mem"])
log_dict["perf_gap(%)"] = calculation_gap(
paddle_num=log_dict["paddle_avg_cost"],
onnxruntime_num=log_dict["onnxruntime_avg_cost"])
log_dict["paddle2onnx_model_convert"] = "Failed"
for model2onnx_success_log in model2onnx_name_list:
if log_dict["model_name"] in model2onnx_success_log:
log_dict[
"paddle2onnx_model_convert"] = model2onnx_success_log.split(
"convert")[-1]
break
else:
continue
data_merging_logs.append(log_dict)
return data_merging_logs
def calculation_gap(paddle_num, onnxruntime_num):
if float(paddle_num) <= 0 and float(onnxruntime_num) <= 0:
return 0
else:
percent = (float(paddle_num) - float(onnxruntime_num)
) / float(onnxruntime_num)
return format(percent * 100, '.3f')
def main(args, result_path, tipc_benchmark_excel_path):
"""
main
"""
# create empty DataFrame
env = args.docker_name
paddle_commit = paddle.__git_commit__
paddle_tag = paddle.__version__
paddle_version = paddle_tag + "/" + paddle_commit
onnxruntime_version = onnxruntime.__version__
origin_df = pd.DataFrame(columns=[
"日期", "环境", "paddle_version", "onnxruntime_version", "device_name",
"model_name", "precision", "batch_size", "enable_mkldnn",
"cpu_threads", "enable_gpu", "enable_trt", "paddle2onnx_model_convert",
"onnxruntime_cpu_mem", "paddle_cpu_mem", "onnxruntime_gpu_mem",
"paddle_gpu_mem", "onnxruntime_avg_cost", "paddle_avg_cost",
"cpu_mem_gap(%)", "gpu_mem_gap(%)", "perf_gap(%)"
])
log_list = log_split(result_path)
model2onnx_name_list = model2onnx_log(result_path)
dict_list = []
for one_model_log in log_list:
output_total_list = process_log(one_model_log)
if "model_name" in output_total_list.keys():
dict_list.append(output_total_list)
output_excl_list = data_merging(env, paddle_version, onnxruntime_version,
model2onnx_name_list, dict_list)
for one_log in output_excl_list:
origin_df = origin_df.append(one_log, ignore_index=True)
raw_df = origin_df.sort_values(by='device_name')
raw_df.to_excel(tipc_benchmark_excel_path)
if __name__ == "__main__":
args = parse_args()
result_path = args.result_path
tipc_benchmark_excel_path = args.output_name
main(args, result_path, tipc_benchmark_excel_path)