-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
executable file
·195 lines (169 loc) · 7.6 KB
/
train.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
#%%
from _0_preprocessing import train_loader,val_loader,test_loader #, x_test, y_test
import torch as tc
import gc
class Relu:
def forward(X): return tc.where(X>0,X,0)
def backward(X): return tc.where(X>0,1,0)
class Conv:
def __init__(self,f,p,s,n_C_prev,n_C_now):
# filter size | padding | stride
self.f, self.p, self.s = f, p, s
self.n_C_prev, self.n_C_now = n_C_prev, n_C_now
beta = 0.1
self.F = tc.rand(n_C_now,n_C_prev,f,f)*beta*2 - beta # -range ~ range # -0.1 ~ 0.1 (0~0.2 - 0.1)
self.b = tc.rand(1,n_C_now,1,1) *beta*2 - beta
self.dF,self.db = tc.zeros_like(self.F),tc.zeros_like(self.b)
def forward(self,A_prev):
self.A_prev = A_prev
self.Z = tc.conv2d(input=A_prev,weight=self.F,stride=self.s,padding=self.p) + self.b
return self.Z
def backward(self, dZ):
self.dF = tc.conv2d(input = self.A_prev.permute(1,0,2,3), weight = dZ.permute(1,0,2,3), stride = self.s,padding=self.p).permute(1,0,2,3)
self.db = tc.sum(dZ,axis = [0,2,3]).reshape(1,self.n_C_now,1,1) # 合理
dA_prev = tc.conv_transpose2d(input = dZ,weight = self.F,stride=self.s,padding=self.p)
return dA_prev
class BatchNorm:
def __init__(self) -> None:
pass
def forward(self):
return
def backward(self):
return
class Pool:
def __init__(self,f,s) -> None:
self.f, self.s = f, s
self.pool = tc.nn.MaxPool2d(kernel_size=f,stride=s,padding=0,return_indices=True)
self.unPool = tc.nn.MaxUnpool2d(kernel_size=f,stride=s,padding=0)
def forward(self,A_prev): #!! should be Z_prev
self.A_prev_size = A_prev.size()
A, self.A_index = self.pool(A_prev)
return A
def backward(self,dA):
dA_prev = self.unPool(dA,self.A_index,output_size = self.A_prev_size)
return dA_prev
class Fc:
def __init__(self,n_prev,n_now) -> None:
self.n_prev, self.n_now = n_prev, n_now
beta = 0.1
self.W = tc.rand(n_now,n_prev)*beta*2 - beta # -range ~ range # -0.1 ~ 0.1
self.b = tc.rand(n_now,1) *beta*2 - beta
self.dW, self.db = tc.zeros_like(self.W),tc.zeros_like(self.b)
def forward(self, A_prev):
self.A_prev = A_prev
Z = tc.mm(self.W,A_prev) + self.b
return Z
def backward(self,dZ):
self.dW = tc.mm(dZ,self.A_prev.T) # (n_now,m)(m,n_prev) = (n_now,n_prev)
self.db = tc.sum(dZ,axis = 1).reshape(dZ.shape[0],1) # (n_now,1)
dA_prev = tc.mm((self.W).T,dZ)
return dA_prev
def forward_prop(X,ls):
Z0 = ls['conv0'].forward(X) # 0: conv0
A0 = Relu.forward(Z0)
A1 = ls['pool1'].forward(A0) # 1: pool0
Z2 = ls['conv2'].forward(A1) # 2: conv1
A2 = Relu.forward(Z2)
A3 = ls['pool3'].forward(A2)
Z4 = ls['conv4'].forward(A3) # Z0 = F0 A_prev0 + b0
A4 = Relu.forward(Z4)
A5 = ls['pool5'].forward(A4) # A5 (m, ch_now, f_w, f_h)
#---------- 後段NN架構 ------------
m = A5.shape[0]
A5_new = A5.reshape(m,-1).T #(ch_now*f_w*f_h, m)
Z6 = ls['fc6'].forward(A5_new)
Aout = tc.nn.functional.softmax(Z6,dim=0)
return Aout
def backward_prop(Aout,ls,Y_hot):
m = Y_hot.shape[1]
#---------- 後段NN架構 ------------
dZ6 = 1/m * (Aout - Y_hot)
dA5_new = ls['fc6'].backward(dZ6)
dA5 = dA5_new.T.reshape(m,dA5_new.shape[0],1,1)
dA4 = ls['pool5'].backward(dA5) #dA2(m,16,10,10)
dZ4 = dA4*Relu.backward(ls['conv4'].Z) #dZ2(m,10,11,11)
ls['conv4'].Z = 0
dA3 = ls['conv4'].backward(dZ4)
dA2 = ls['pool3'].backward(dA3) #dA2(m,16,10,10)
dZ2 = dA2*Relu.backward(ls['conv2'].Z) #dZ2(m,10,11,11)
ls['conv2'].Z = 0
dA1 = ls['conv2'].backward(dZ2)
dA0 = ls['pool1'].backward(dA1) #dA1(m,6,28,28)
dZ0 = dA0*Relu.backward(ls['conv0'].Z) #dZ1(m,6,28,28)
ls['conv0'].Z = 0
_ = ls['conv0'].backward(dZ0)
return
def update_params(alpha, ls,m,reg):
for key in ls:
if(type(ls[key]) == Fc):
ls[key].W = ls[key].W*(1-alpha*reg/m) - alpha * ls[key].dW
ls[key].b = ls[key].b*(1-alpha*reg/m) - alpha * ls[key].db
elif(type(ls[key]) == Conv):
ls[key].F = ls[key].F*(1-alpha*reg/m) - alpha * ls[key].dF
ls[key].b = ls[key].b*(1-alpha*reg/m) - alpha * ls[key].db
return
from utils.gpu import gpu_acceleration
loss_list = []
acc_list = {"train":[],"valid":[],"test":[]}
def get_accu(Aout,y_labels,data_type,iteration,acc_list):
A_pred = tc.argmax(Aout,axis = 0)
acc_num, tot= tc.sum(A_pred.cpu() == y_labels), len(y_labels)
print(f"train acc {(acc_num/tot).item()*100:.1f} {acc_num}/{tot}")
acc_list[data_type].append([iteration,(acc_num/tot).item()])
return
def start_training(train_loader,ls,epochs,lr,reg=3):
for key in ls:
if(type(ls[key]) == Fc): ls[key].W, ls[key].b = gpu_acceleration(ls[key].W, ls[key].b)
elif(type(ls[key]) == Conv): ls[key].F, ls[key].b = gpu_acceleration(ls[key].F, ls[key].b)
iteration = 0
for epoch in range(epochs):
for i, (X,y_labels) in enumerate(train_loader):
# print("iter", i)
iteration+=1
m = X.shape[0]
Y_hot = tc.nn.functional.one_hot(y_labels,num_classes=10).T #Y_hot: (10,m)
X,Y_hot = gpu_acceleration(X,Y_hot)
Aout = forward_prop(X,ls)
backward_prop(Aout,ls,Y_hot)
update_params(lr, ls,m,reg)
if ((epoch % 1) == 0 or (epoch == epochs-1)) and (i%20==0):
print(f"-----------epoch {epoch}/{epochs}---i {i}/{len(train_loader)}------")
loss = - tc.sum(Y_hot * tc.log(Aout))/Aout.shape[1]
# print(f"loss {loss.cpu().item()}")
loss_list.append([iteration,loss.cpu().item()])
get_accu(Aout,y_labels,"train",iteration,acc_list)
X,y_labels = iter(val_loader).next()
Aout = forward_prop(X.cuda(),ls)
get_accu(Aout,y_labels,"valid",iteration,acc_list)
X,y_labels = iter(test_loader).next()
Aout = forward_prop(X.cuda(),ls)
get_accu(Aout,y_labels,"test",iteration,acc_list)
# print(F1.shape,F2.shape)
return
if __name__ == "__main__":
layer_list = {
'conv0': Conv(f=5,p=2,s=1,n_C_prev=3,n_C_now=32), #? layer0 (m,3,32,32) --> (m,9,32,32)
'pool1': Pool(f=2,s=2), # layer1 (m,6,32,32) --> (m,6,16,16)
'conv2': Conv(f=5,p=0,s=1,n_C_prev=32,n_C_now=64), #? layer2 (m,6,16,16) --> (m,16,12,12)
'pool3': Pool(f=2,s=2), # layer3 (m,16,12,12) --> (m,16,6,6)
'conv4': Conv(f=5,p=0,s=1,n_C_prev=64,n_C_now=128),#layer (m,16,6,6) --> (m,32,2,2)
'pool5': Pool(f=2,s=2),# layer5 (m,32,2,2) --> (m,32,1,1)
'fc6' : Fc(n_prev=128,n_now=10), #* layer6 (32,m) --> (10,m)
}
import time, os, pickle
start = time.time() # 25
start_training(train_loader,layer_list,epochs=3,lr=0.10,reg=0)
end = time.time()
duration = int(end - start)
print("duration",duration)
cnnNetInfo = {
'file': 'v2_modulization',
'loss_list': loss_list, 'acc_list': acc_list,
'training_time(s)': duration,
'layerList': layer_list
}
outdir = './_train_ML_data/train_W_b_result'
if not os.path.exists(outdir): os.makedirs(outdir)
with open(f"{outdir}/v2_modulization.pkl",'wb') as cnnInfoPickle:
pickle.dump(cnnNetInfo,cnnInfoPickle)
print("Piiiiiclke")