-
Notifications
You must be signed in to change notification settings - Fork 7
/
generate_rt_ai_model_h.py
140 lines (106 loc) · 4.82 KB
/
generate_rt_ai_model_h.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
# coding=utf-8
'''
@ Summary: based on {X-CUBE-AI/<model>.h, X-CUBE-AI/<model>_data.h, Documents/rt_ai_mnist_model.h}
to generate <project>/applications/rt_ai_<model>_model.h
@ Update:
@ file: generate_rt_ai_model_h.py
@ version: 1.0.0
@ Author: Lebhoryi@gmail.com
@ Date: 2021/1/29 14:12
'''
import os
import sys
import logging
from pathlib import Path
path = os.path.dirname(__file__)
sys.path.append(os.path.join(path, '../../'))
def check_file(file):
assert Path(file).exists(), logging.error(f"No such file {file} exists!")
def read_file(file):
check_file(file)
with open(file, "r") as f:
lines = f.readlines()
return lines
def get_model_info(template, model_h, model_name, default_name="network"):
# model_info = [] # modol informations
input_nums, output_nums = 1, 1
template_lines = read_file(template)
model_h_lines = read_file(model_h)
for line in model_h_lines:
if "AI_NETWORK_IN_NUM" in line:
input_nums = int(line.split()[-1][1:-1])
elif "AI_NETWORK_OUT_NUM" in line:
output_nums = int(line.split()[-1][1:-1])
all_input_size_bytes = [f"AI_{model_name.upper()}_IN_{i+1}_SIZE_BYTES" for i in range(input_nums)]
all_output_size_bytes = [f"AI_{model_name.upper()}_OUT_{i+1}_SIZE_BYTES" for i in range(output_nums)]
for i, line in enumerate(template_lines):
# replace model name
if model_name != default_name and default_name in line:
template_lines[i] = template_lines[i].replace(default_name, model_name)
if model_name != default_name and default_name.upper() in line:
template_lines[i] = template_lines[i].replace(default_name.upper(), model_name.upper())
if "IN_TOTAL_SIZE_BYTES" in line:
in_total_size_bytes = "(" + "+".join(all_input_size_bytes) + ")"
template_lines[i] = template_lines[i].replace("NULL", in_total_size_bytes)
elif "IN_SIZE_BYTES" in line:
in_size_bytes = "{" + ", ".join(all_input_size_bytes) + "}"
template_lines[i] = template_lines[i].replace("NULL", in_size_bytes)
elif "OUT_TOTAL_SIZE_BYTES" in line:
out_total_size_bytes = "(" + "+".join(all_output_size_bytes) + ")"
template_lines[i] = template_lines[i].replace("NULL", out_total_size_bytes)
elif "OUT_SIZE_BYTES" in line:
out_size_bytes = "{" + ", ".join(all_output_size_bytes) + "}"
template_lines[i] = template_lines[i].replace("NULL", out_size_bytes)
return template_lines
def rt_ai_model_gen(stm_out, project, model_name, rt_ai_example):
''' generate rt_ai_<model_name>_model.h'''
# where save <model_name>.h and <model_name>_data.h
x_cube_ai = Path(stm_out) / "X-CUBE-AI/App"
assert x_cube_ai.exists(), "No X-CUBE-AI/App exists, pls check the path!!!"
# the files from x-cube-ai
model_h = x_cube_ai / (model_name + ".h")
# modol informations
template_file = Path(rt_ai_example)/"rt_ai_template_model.h"
model_info = get_model_info(template_file, model_h, model_name)
# project/applications/<model>.h
pro_app_model_h = Path(project) / "applications" / f"rt_ai_{model_name}_model.h"
if pro_app_model_h.exists(): pro_app_model_h.unlink()
with pro_app_model_h.open("w+") as fw:
fw.write("".join(model_info))
logging.info(f"Generate rt_ai_{model_name}_model.h successfully...")
if __name__ == "__main__":
from prepare_work import pre_sconscript
from plugin_init import set_env
from run_x_cube_ai import stm32ai
logging.getLogger().setLevel(logging.INFO)
# 1. set env
ext_tools = r"D:\Program Files (x86)\stm32ai-windows-7.0.0\windows"
_ = set_env(ext_tools)
# 2. prepare tmp output
aitools_out, dir_names, scons_path = 'tmp_cwd', ["Middlewares", "X-CUBE-AI"], "Sconscripts"
_ = pre_sconscript(aitools_out, scons_path, dir_names)
# 3. test stm32ai
class Opt():
def __init__(self):
self.stm_out = "tmp_cwd/stm32ai_middle"
self.model_name = "network"
self.workspace = "./stm32ai_ws"
self.compress = 1
self.batches = 10
self.mode = "011"
self.val_data = ''
self.cube_ai = r"D:\Program Files (x86)\Keil_v5\PACK\STMicroelectronics\X-CUBE-AI\7.0.0"
opt = Opt()
model = "../../Models/mnist.tflite"
stm_out = "tmp_cwd"
c_model_name = "network"
sup_modes = ["001", "011", "101", "111"]
ai_params = [opt.workspace, opt.compress, opt.batches, opt.mode, opt.val_data]
_ = stm32ai(model, stm_out, c_model_name, sup_modes, ai_params)
project = Path("tmp_cwd")
tmp_project = project / "applications"
if not tmp_project.exists():
tmp_project.mkdir()
rt_ai_example = "../../Documents"
_ = rt_ai_model_gen(stm_out, project, "network")
print("u a right...")