-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtransfer.py
142 lines (129 loc) · 5.89 KB
/
transfer.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
# import torch
import paddle
from collections import OrderedDict
def transfer():#transfer linear
input_fp = "yolox_s.pth"
output_fp = "yolox_s.pdparams"
torch_dict = torch.load(input_fp,map_location=torch.device('cpu'))['model']
paddle_dict ={}
fc_names = []
for key in torch_dict:
weight = torch_dict[key].cpu().numpy()
flag = [i in key for i in fc_names]
if any(flag):
print("weight{}need to be tansferd".format(key))
weight = weight.transpose()
paddle_dict[key] = weight
paddle.save(paddle_dict,output_fp)
def show_keys_torch():
params = torch.load("yolox_s.pth", map_location=torch.device('cpu'))['model']
for key, value in params.items():
print(key,value.shape)
def show_keys_paddle():
# params=paddle.load("yolox_s.pdparams")
params=paddle.load("yolox_new02.pdparams")
for key, value in params.items():
print(key,value.shape)
def show_value_paddle():
params=paddle.load("yolox_s.pdparams")
for key, value in params.items():
print(key,value)
def switch_keys():#用于转换排查出不同名字的权重
new_params=paddle.load("yolo.pdparams")
params=paddle.load("yolox_s.pdparams")
for k, v in params.items():
if k.endswith(".running_mean"):
new_params[k.replace(".running_mean","._mean")] = v.detach().numpy()
elif k.endswith(".running_var"):
new_params[k.replace(".running_var","._variance")] = v.detach().numpy()
bn_w_name_list.append(k.replace(".running_var", ".weight"))
else:
new_params[k] = v.detach().numpy()
paddle_model.set_dict(new_params)
#用于排查出不同名字的权重值转换
def switch_values():
paddle_weight=paddle.load("yoloQAQ.pdparams")
p_weight = paddle.load("yolox_s.pdparams")
paddle_keys = [] # 存放paddle模型的权重键值
p_keys = [] # 存放torch模型的权重键值
for k in p_weight: # 遍历torch模型权重键值
p_keys.append(k)
for k in paddle_weight: # 遍历paddle模型权重键值
paddle_keys.append(k)
key_pair_length = min(len(p_keys), len(paddle_keys)) # 获取最小对应权重长度
# 将pytorch模型参数赋值给paddle模型
for i, k in enumerate(paddle_keys):
if i >= key_pair_length:
break
if paddle_weight[k].shape == p_weight[p_keys[i]].detach().numpy().shape: # 权重参数shape比较，只有一一对应才会赋值
paddle_weight[k] = p_weight[p_keys[i]].detach().numpy()
for k, p in p_weight.items():
if k in paddle_weight:
p_param = p_weight[k].detach().numpy()
if p_param.shape == paddle_weight[k].shape:
paddle_weight[k] = p_param
else:
print('torch param {} dose not match paddle param {}'.format(k, k))
elif 'running_mean' in k:
p_param = p_weight[k].detach().numpy()
if p_param.shape == paddle_weight[k[:-12]+'_mean'].shape:
paddle_weight[k[:-12]+'_mean'] = p_param
else:
print('torch param {} dose not match paddle param {}'.format(k, k[:-12]+'_mean'))
elif 'running_var' in k:
p_param = p_weight[k].detach().numpy()
if p_param.shape == paddle_weight[k[:-11] + '_variance'].shape:
paddle_weight[k[:-11] + '_variance'] = p_param
else:
print('torch param {} dose not match paddle param {}'.format(k, k[:-11] + '_variance'))
else:
print('torch param {} not exist in paddle modle'.format(k))
def show_value():
# params1=paddle.load("yolox_s_new0.pdparams")
params1=paddle.load("_coco_eval_log.pdparams")
print("paddle")
for key1, value1 in params1.items():
if 'head' in key1:
print(key1,value1.shape)
# if '.reg_convs.5.bn._variance ' in key1:
# print(key1,value1)
# if key1.endswith(".reg_convs.5.bn._variance "):
# print(key1,value1)
# if key1.endswith(".reg_convs.5.bn._variance "):
# print(key1,value1)
def change_value():
torch_keys = [] # 存放torch模型的权重键值
paddle_keys = [] # 存放paddle模型的权重键值
torch_weight = torch.load(input_fp,map_location=torch.device('cpu'))['model']
paddle_weight=paddle.load("yolox_s.pdparams")
for k in torch_weight.items(): # 遍历torch模型权重键值
torch_keys.append(k)
for k in paddle_weight.items(): # 遍历paddle模型权重键值
paddle_keys.append(k)
key_pair_length = min(len(torch_keys), len(paddle_keys)) # 获取最小对应权重长度
# 将pytorch模型参数赋值给paddle模型
for i, k in enumerate(paddle_keys):
if i >= key_pair_length:
break
if paddle_weight[k].shape == torch_weight[torch_keys[i]].detach().numpy().shape: # 权重参数shape比较，只有一一对应才会赋值
paddle_weight[k] = torch_weight[torch_keys[i]].detach().numpy()
# 将paddle模型参数赋值给pytorch模型
for i, k in enumerate(torch_keys):
if i >= key_pair_length:
break
if torch_weight[k].detach().numpy().shape == paddle_weight[paddle_keys[i]].shape: # 权重参数shape比较，只有一一对应才会赋值
torch_weight[k] = paddle_weight[paddle_keys[i]]
def show():
paddle_weight=paddle.load("yoloQAQ.pdparams")
for k, p in paddle_weight.items():
if k in paddle_weight:
param = paddle_weight[k].detach().numpy()
if param.shape == paddle_weight[k].shape:
paddle_weight[k] = p_param
#head.reg_convs.5.bn._variance
if __name__ == "__main__":
# transfer()
# switch_values()
# show_keys_paddle()
# show_keys_torch()
show_value()