-
Notifications
You must be signed in to change notification settings - Fork 20
/
performance_accuracy.py
393 lines (300 loc) · 16.9 KB
/
performance_accuracy.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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
# ====TensorRT Hackathon 2022 Medcare AI Lab====
#
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
#
'''
MST++ TensorRT 绝对误差和相对误差的平均值，中位数，最大值计算和可视化
'''
import os
import argparse
import ctypes
from glob import glob
import architecture
import torch
import onnxruntime
import numpy as np
import heapq
import pandas as pd
import tensorrt as trt
from cuda import cudart
import cv2
from time import time_ns,sleep
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib
matplotlib.use('Agg')
plt.style.use('ggplot')
files = os.listdir("./data")
soFileList = glob("./" + "*.so")
#pytorch
torch_model = architecture.MST_Plus_Plus() # 导入模型
check_point = torch.load("./model/mst_plus_plus.pth",map_location='cpu')
# torch_model.load_state_dict(check_point['state_dict']) # 初始化权重
torch_model.load_state_dict({k.replace('module.', ''): v for k, v in check_point['state_dict'].items()},
strict=True)
# torch_model.half()
torch_model.eval()
torch_model.to("cuda")
# trt
logger = trt.Logger(trt.Logger.VERBOSE) # trt.Logger.VERBOSE
trt.init_libnvinfer_plugins(logger, '')
PLUGIN_CREATORS = trt.get_plugin_registry().plugin_creator_list
for soFile in soFileList:
ctypes.cdll.LoadLibrary(soFile)
EXPLICIT_BATCH = 1 << (int)(trt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH)
builder = trt.Builder(logger)
network = builder.create_network(EXPLICIT_BATCH)
config = builder.create_builder_config()
config.max_workspace_size = 1 << 30
#load engine
if os.path.exists("./model/mst_plus_plus_b1.plan"):
# If a serialized engine exists, load it instead of building a new one.
with open("./model/mst_plus_plus_b1.plan", "rb") as f:
engine = trt.Runtime(logger).deserialize_cuda_engine(f.read())
if engine == None:
print("Failed loading %s"%decoderPlanFile)
exit()
context = engine.create_execution_context()
#load engine
if os.path.exists("./model/mst_plus_plus_onnxparser_b1.plan"):
# If a serialized engine exists, load it instead of building a new one.
with open("./model/mst_plus_plus_onnxparser_b1.plan", "rb") as f:
engine_1 = trt.Runtime(logger).deserialize_cuda_engine(f.read())
if engine_1 == None:
print("Failed loading %s"%decoderPlanFile)
exit()
context_1 = engine_1.create_execution_context()
if os.path.exists("./model/mst_plus_plus_b1_fp16.plan"):
# If a serialized engine exists, load it instead of building a new one.
with open("./model/mst_plus_plus_b1_fp16.plan", "rb") as f:
engine_2 = trt.Runtime(logger).deserialize_cuda_engine(f.read())
if engine_2 == None:
print("Failed loading %s"%decoderPlanFile)
exit()
context_2 = engine_2.create_execution_context()
# load engien 2
#load engine
if os.path.exists("./model/mst_plus_plus_onnxparser_b1_fp16.plan"):
# If a serialized engine exists, load it instead of building a new one.
with open("./model/mst_plus_plus_onnxparser_b1_fp16.plan", "rb") as f:
engine_3 = trt.Runtime(logger).deserialize_cuda_engine(f.read())
if engine_3 == None:
print("Failed loading %s"%decoderPlanFile)
exit()
context_3 = engine_3.create_execution_context()
#onnxruntime
ort_session = onnxruntime.InferenceSession("./model/mst_plus_plus_b1.onnx",providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
# providers=['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
def acc_static_torch(input,model=torch_model):
torch_input = torch.from_numpy(input).cuda()
with torch.no_grad():
torch_output = model(torch_input)
return torch_output.cpu().detach().numpy()[:, :, 128:-128, 128:-128]
def acc_static_onnx(input,model=ort_session):
ort_inputs = {"input":input}
output_name = ort_session.get_outputs()[0].name
output = model.run([output_name],ort_inputs)
return output[0][:, :, 128:-128, 128:-128]
def acc_static_trt(input,engine,context):
inputH0 = np.ascontiguousarray(input.ravel())
_, inputD0 = cudart.cudaMalloc(inputH0.nbytes)
cudart.cudaMemcpy(inputD0, inputH0.ctypes.data, inputH0.nbytes, cudart.cudaMemcpyKind.cudaMemcpyHostToDevice)
outputH0 = np.empty(context.get_binding_shape(1),dtype=trt.nptype(engine.get_binding_dtype(1)))
_, outputD0 = cudart.cudaMalloc(outputH0.nbytes)
context.execute_v2([int(inputD0),int(outputD0)])
cudart.cudaMemcpy(outputH0.ctypes.data, outputD0, outputH0.nbytes, cudart.cudaMemcpyKind.cudaMemcpyDeviceToHost)
cudart.cudaFree(inputD0)
cudart.cudaFree(outputD0)
return outputH0[:, :, 128:-128, 128:-128]
def cut_dat(x,n=20):
'''
为了保持统计的合理性，去掉前5个最大值和最小值在做统计
'''
x_index = map(x.index,heapq.nlargest(n, x))
y_index = map(x.index,heapq.nsmallest(n, x))
for i in list(x_index)+list(y_index):
x[i] = 0.0
return x
if __name__ == "__main__":
# 平均绝对误差的计算
# np.mean(np.abs(output_data_pytorch - output_data_trt)
# 平均相对精度的计算
# np.mean(np.abs(output_data_pytorch - output_data_trt) / np.abs(output_data_pytorch))
# 只在途中画出相对绝对误差的 最大值，最小值，和中位数
files = os.listdir("./data")
abserror_onnx = []
abserror_onnxparser_32 = []
abserror_api_32 = []
abserror_api_16 = []
abserror_onnxparser_16 = []
abserror_onnxparser_16 = []
relerror_onnx = []
relerror_onnxparser_32 = []
relerror_api_32 = []
relerror_onnxparser_16 = []
relerror_api_16 = []
relerror_onnxparser_16 = []
for file in files:
print(file)
file_path = os.path.join("./data/",file)
img = cv2.imread(file_path)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_fp32 = np.array(img,dtype=np.float32)
img_fp32 = (img_fp32 - img_fp32.min()) / (img_fp32.max() - img_fp32.min())
img_1 = img_fp32.transpose([2,1,0]).reshape(1,3,512,482) # 3x512x482
torch_output = acc_static_torch(img_1)
onnx_output = acc_static_onnx(img_1)
onnxparser_32_output = acc_static_trt(img_1,engine_1,context_1)
onnxparser_16_output = acc_static_trt(img_1,engine_3,context_3)
api_32_output = acc_static_trt(img_1,engine,context)
api_16_output = acc_static_trt(img_1,engine_2,context_2)
# 绝对误差 计算指标
abserror_onnx.append( np.mean(np.abs(torch_output-onnx_output)) )
abserror_onnxparser_32.append( np.mean(np.abs(torch_output-onnxparser_32_output)) )
abserror_api_32.append( np.mean(np.abs(torch_output-api_32_output)) )
abserror_onnxparser_16.append( np.mean(np.abs(torch_output-onnxparser_16_output)) )
abserror_api_16.append( np.mean(np.abs(torch_output-api_16_output)) )
# 相对误差 计算指标
relerror_onnx.append( np.mean(np.abs(torch_output-onnx_output) / (np.abs(torch_output)+1e-12)) )
relerror_onnxparser_32.append( np.mean(np.abs(torch_output-onnxparser_32_output) /(np.abs(torch_output)+1e-12) ) )
relerror_api_32.append( np.mean(np.abs(torch_output-api_32_output) /(np.abs(torch_output)+1e-12) ) )
relerror_onnxparser_16.append( np.mean(np.abs(torch_output-onnxparser_16_output) /(np.abs(torch_output)+1e-12) ) )
relerror_api_16.append( np.mean(np.abs(torch_output-api_16_output) /(np.abs(torch_output)+1e-12) ) )
# Category = ["onnxruntime"] * (len(abserror_onnx))+["TensorRT ONNXParser(FP32)"] * len(abserror_onnxparser_32) + ["TensorRT API(FP32)"]*len(abserror_api_32) +["TensorRT ONNXParser(FP16)"] * len(abserror_onnxparser_16) + ["TensorRT API(FP16)"] * len(abserror_api_16)
# abs_error = {'Category':Category,
# 'Absolute Error':cut_dat(abserror_onnx) + cut_dat(abserror_onnxparser_32) + cut_dat(abserror_api_32) + cut_dat(abserror_onnxparser_16)+cut_dat(abserror_api_16),
# 'File Index': list(range(len(files)))*5}
# df = pd.DataFrame(abs_error)
# sns.jointplot(data=df, x='File Index', y='Absolute Error', hue='Category')
# col_labels = ['Mean','Median','Max']
# row_labels = ['onnxruntime','ONNXParser(FP32)','TensorRT API(FP32)','ONNXParser(FP16)','TensorRT API(FP16)']
# table_vals = [[np.mean(cut_dat(abserror_onnx)),np.max(cut_dat(abserror_onnx)),np.min(cut_dat(abserror_onnx))],
# [round(np.mean(cut_dat(abserror_onnxparser_32)),9),np.max(cut_dat(abserror_onnxparser_32)),np.min(cut_dat(abserror_onnxparser_32))],
# [round(np.mean(cut_dat(abserror_api_32)),9),np.max(cut_dat(abserror_api_32)),np.min(cut_dat(abserror_api_32))],
# [round(np.mean(cut_dat(abserror_onnxparser_16)),9),np.max(cut_dat(abserror_onnxparser_16)),np.min(cut_dat(abserror_onnxparser_16))],
# [round(np.mean(cut_dat(abserror_api_16)),9),np.max(cut_dat(abserror_api_16)),np.min(cut_dat(abserror_api_16))]]
# row_colors = ['red','gold','green',"blue","purple"]
# table = plt.table(cellText=table_vals, #colWidths=[0.1]*8,
# rowLabels=row_labels, colLabels=col_labels,
# rowColours=row_colors, colColours=row_colors,
# bbox=(-3, 0.6, 2, 0.2))
# table.auto_set_font_size(False)
# table.set_fontsize(10)
# plt.savefig("./mean_abs_error.png",bbox_inches='tight', pad_inches=0)
# plt.close()
# mean abs
plt.rcParams['figure.figsize'] = (16.0, 9.0) # 单位是inches
fig,subs=plt.subplots(2,3)
subs[0][0].plot(np.arange(len(files)), cut_dat(abserror_onnx), 'ro',label='onnxruntime')
# subs[0][0].axhline(y=np.max(mean_abserror_onnx),color='b',linestyle='--')
# subs[0][0].axhline(y=np.min(mean_abserror_onnx),color='b',linestyle='--')
# subs[0][0].axhline(y=np.min(mean_abserror_onnx),color='b',linestyle='--')
subs[0][0].set_xlabel('Test Image ID')
subs[0][0].set_ylabel("Absolute Error")
subs[0][0].legend(edgecolor='blue')
subs[0][1].plot(np.arange(len(files)), cut_dat(abserror_onnxparser_32), 'bv',label='TensorRT ONNXParser (FP32)')
# subs[0][1].axhline(y=np.max(mean_abserror_onnxparser_32),color='b',linestyle='--')
# subs[0][1].axhline(y=np.min(mean_abserror_onnxparser_32),color='b',linestyle='--')
subs[0][1].set_xlabel('Test Image ID')
subs[0][1].set_ylabel("Absolute Error")
subs[0][1].legend(edgecolor='blue')
subs[0][2].plot(np.arange(len(files)), cut_dat(abserror_api_32), 'g^',label='TensorRT API (FP32)')
# subs[0][2].axhline(y=np.max(mean_abserror_api_32),color='b',linestyle='--')
# subs[0][2].axhline(y=np.min(mean_abserror_api_32),color='b',linestyle='--')
subs[0][2].set_xlabel('Test Image ID')
subs[0][2].set_ylabel("Absolute Error")
subs[0][2].legend(edgecolor='blue')
subs[1][0].plot(np.arange(len(files)), cut_dat(abserror_onnxparser_16), 'k*',label='TensorRT ONNXParser (FP16)')
# subs[1][0].axhline(y=np.max( mean_abserror_onnxparser_16),color='b',linestyle='--')
# subs[1][0].axhline(y=np.min( mean_abserror_onnxparser_16),color='b',linestyle='--')
subs[1][0].set_xlabel('Test Image ID')
subs[1][0].set_ylabel("Absolute Error")
subs[1][0].legend(edgecolor='blue')
subs[1][1].plot(np.arange(len(files)), cut_dat(abserror_api_16), 'mh',label='TensorRT API (FP16)')
# subs[1][1].axhline(y=np.max(mean_abserror_api_16),color='b',linestyle='--')
# subs[1][1].axhline(y=np.min(mean_abserror_api_16),color='b',linestyle='--')
subs[1][1].set_xlabel('Test Image ID')
subs[1][1].set_ylabel("Absolute Error")
subs[1][1].legend(edgecolor='blue')
# plt.delaxes(subs[1][2])
plt.delaxes()
col_labels = ['Mean','Median','Max']
row_labels = ['onnxruntime','ONNXParser(FP32)','TensorRT API(FP32)','ONNXParser(FP16)','TensorRT API(FP16)']
table_vals = [[round(np.mean(cut_dat(abserror_onnx)),8),round(np.median(cut_dat(abserror_onnx)),8),round(np.max(cut_dat(abserror_onnx)),8)],
[round(np.mean(cut_dat(abserror_onnxparser_32)),8),round(np.median(cut_dat(abserror_onnxparser_32)),8),round(np.max(cut_dat(abserror_onnxparser_32)),8)],
[round(np.mean(cut_dat(abserror_api_32)),8),round(np.median(cut_dat(abserror_api_32)),8),round(np.max(cut_dat(abserror_api_32)),8)],
[round(np.mean(cut_dat(abserror_onnxparser_16)),8),round(np.median(cut_dat(abserror_onnxparser_16)),8),round(np.max(cut_dat(abserror_onnxparser_16)),8)],
[round(np.mean(cut_dat(abserror_api_16)),8),round(np.median(cut_dat(abserror_api_16)),8),round(np.max(cut_dat(abserror_api_16)),8)]]
row_colors = ['red','gold','green',"blue","purple"]
table = subs[1][1].table(cellText=table_vals, colWidths=[0.1]*3,
rowLabels=row_labels, colLabels=col_labels,
rowColours=row_colors, colColours=row_colors,
bbox=(-0.75, -0.7, 2.9, 0.5))
table.auto_set_font_size(False)
table.set_fontsize(13)
plt.savefig("./abs_error.png",bbox_inches='tight', pad_inches=0)
plt.close()
# mean relative abs
plt.rcParams['figure.figsize'] = (16.0, 9.0) # 单位是inches
fig,subs=plt.subplots(2,3)
subs[0][0].plot(np.arange(len(files)), cut_dat(relerror_onnx), 'ro',label='onnxruntime')
# subs[0][0].axhline(y=np.max(mean_abserror_onnx),color='b',linestyle='--')
# subs[0][0].axhline(y=np.min(mean_abserror_onnx),color='b',linestyle='--')
# subs[0][0].axhline(y=np.min(mean_abserror_onnx),color='b',linestyle='--')
subs[0][0].set_xlabel('Test Image ID')
subs[0][0].set_ylabel("Relative Error")
subs[0][0].legend(edgecolor='blue')
subs[0][1].plot(np.arange(len(files)), cut_dat(relerror_onnxparser_32), 'bv',label='TensorRT ONNXParser (FP32)')
# subs[0][1].axhline(y=np.max(mean_abserror_onnxparser_32),color='b',linestyle='--')
# subs[0][1].axhline(y=np.min(mean_abserror_onnxparser_32),color='b',linestyle='--')
subs[0][1].set_xlabel('Test Image ID')
subs[0][1].set_ylabel("Relative Error")
subs[0][1].legend(edgecolor='blue')
subs[0][2].plot(np.arange(len(files)), cut_dat(relerror_api_32), 'g^',label='TensorRT API (FP32)')
# subs[0][2].axhline(y=np.max(mean_abserror_api_32),color='b',linestyle='--')
# subs[0][2].axhline(y=np.min(mean_abserror_api_32),color='b',linestyle='--')
subs[0][2].set_xlabel('Test Image ID')
subs[0][2].set_ylabel("Relative Error")
subs[0][2].legend(edgecolor='blue')
subs[1][0].plot(np.arange(len(files)), cut_dat(relerror_onnxparser_16), 'k*',label='TensorRT ONNXParser (FP16)')
# subs[1][0].axhline(y=np.max( mean_abserror_onnxparser_16),color='b',linestyle='--')
# subs[1][0].axhline(y=np.min( mean_abserror_onnxparser_16),color='b',linestyle='--')
subs[1][0].set_xlabel('Test Image ID')
subs[1][0].set_ylabel("Relative Error")
subs[1][0].legend(edgecolor='blue')
subs[1][1].plot(np.arange(len(files)), cut_dat(relerror_api_16), 'mh',label='TensorRT API (FP16)')
# subs[1][1].axhline(y=np.max(mean_abserror_api_16),color='b',linestyle='--')
# subs[1][1].axhline(y=np.min(mean_abserror_api_16),color='b',linestyle='--')
subs[1][1].set_xlabel('Test Image ID')
subs[1][1].set_ylabel("Relative Error")
subs[1][1].legend(edgecolor='blue')
# plt.delaxes(subs[1][2])
plt.delaxes()
col_labels = ['Mean','Median','Max']
row_labels = ['onnxruntime','ONNXParser(FP32)','TensorRT API(FP32)','ONNXParser(FP16)','TensorRT API(FP16)']
table_vals = [[round(np.mean(cut_dat(relerror_onnx)),8),round(np.median(cut_dat(relerror_onnx)),8),round(np.max(cut_dat(relerror_onnx)),8)],
[round(np.mean(cut_dat(relerror_onnxparser_32)),8),round(np.median(cut_dat(relerror_onnxparser_32)),8),round(np.max(cut_dat(relerror_onnxparser_32)),8)],
[round(np.mean(cut_dat(relerror_api_32)),8),round(np.median(cut_dat(relerror_api_32)),8),round(np.max(cut_dat(relerror_api_32)),8)],
[round(np.mean(cut_dat(relerror_onnxparser_16)),8),round(np.median(cut_dat(relerror_onnxparser_16)),8),round(np.max(cut_dat(relerror_onnxparser_16)),8)],
[round(np.mean(cut_dat(relerror_api_16)),8),round(np.median(cut_dat(relerror_api_16)),8),round(np.max(cut_dat(relerror_api_16)),8)]]
row_colors = ['red','gold','green',"blue","purple"]
table = subs[1][1].table(cellText=table_vals, colWidths=[0.1]*3,
rowLabels=row_labels, colLabels=col_labels,
rowColours=row_colors, colColours=row_colors,
bbox=(-0.75, -0.7, 2.9, 0.5))
table.auto_set_font_size(False)
table.set_fontsize(13)
plt.savefig("./rel_error.png",bbox_inches='tight', pad_inches=0)
plt.close()