-
Notifications
You must be signed in to change notification settings - Fork 0
/
pretrainedTerminal.py
189 lines (158 loc) · 8.03 KB
/
pretrainedTerminal.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
import torch
import logging
import argparse
import torchvision
import torch.nn as nn
from tqdm import tqdm
import torch.optim as optim
from torchvision import models
import torch.utils.data as data
from torchvision import transforms
def get_args():
parser = argparse.ArgumentParser(description='Modelos pre-entrenados con pytorch')
parser.add_argument('--epochs', '-e', type=int, default=4, help='Numero de epocas')
parser.add_argument('--model', '-m', type=str, default='resnet50', help='Modelo a utilizar')
parser.add_argument('--batch_size', '-bs', type=int, default=32, help='Tamaño del batch')
parser.add_argument('--learning_rate', '-lr', type=float, default=1e-3, help='Learning rate', dest='lr')
parser.add_argument('--re_dim', '-rd', type=int, default=120, help='Re dimensionar las imagenes')
parser.add_argument('--clases', '-c', type=int, default=2, help='Numero de clases')
parser.add_argument('--neurons', '-n', type=int, default=500, help='Numero de capas')
parser.add_argument('--num_cap', '-nc', type=int, default=1, help='Numero de capas')
parser.add_argument('--ent_path', '-ep', type=str,
default= 'DATA_TRAINING',
help='Direccion de los datos de entrenamiento')
parser.add_argument('--val_path', '-cvp', type=str,
default= 'DATA_VALIDATION',
help='Direccion de los datos de validacion')
return parser.parse_args()
def creatingData(args):
trainDataPath = args.ent_path
valDataPath = args.val_path
transforms_images = transforms.Compose([
transforms.Resize((args.re_dim, args.re_dim)),
transforms.ToTensor(),
transforms.Normalize(mean= [0.485, 0.456, 0.406],
std= [0.229, 0.224, 0.225])
])
valData = torchvision.datasets.ImageFolder(root = valDataPath, transform = transforms_images)
trainData = torchvision.datasets.ImageFolder(root= trainDataPath, transform= transforms_images)
trainDataLoader= data.DataLoader(trainData, batch_size= args.batch_size, shuffle=True)
valDataLoader= data.DataLoader(valData, batch_size= args.batch_size, shuffle=True)
return trainDataLoader, valDataLoader
def checkingGPU():
if torch.cuda.is_available():
device = torch.device("cuda")
else:
device = torch.device("cpu")
logging.info(f'Using device {device}')
return device
def getNeurons(nc):
ln = []
for i in range(nc):
ln.append(int(input(f'Layer {i} neurons: ')))
return ln
def trainingModel(model, output= 2, n_layers= 1, n_features= []):
transfer_model = model
layers = []
if n_layers > 1:
layers.append(nn.Linear(transfer_model.fc.in_features, n_features[0]))
for i in range(len(n_features)- 1):
layers.append(nn.Linear(n_features[i], n_features[i + 1]))
else:
layers.append(nn.Linear(transfer_model.fc.in_features, n_features[0]))
transfer_model.fc = nn.Sequential(*layers,
nn.ReLU(),
nn.Dropout(),
nn.Linear(n_features[-1], output),
nn.Softmax(dim= 0))
return transfer_model
def menuDisplay(modelv):
preTrained = {'resnet50': models.resnet50,
'resnet101': models.resnet101,
'resnet152': models.resnet152,
'squeezenet1_0': models.squeezenet1_0,
'squeezenet1_1': models.squeezenet1_1,
'convnext_large': models.convnext_large,
'convnext_small': models.convnext_small,
'inception_v3': models.inception_v3,
'mobilenet_v3_large': models.inception_v3,
'googlenet': models.googlenet,
'efficientnet_b7': models.efficientnet_b7,
'efficientnet_b0': models.efficientnet_b0,
}
if modelv in preTrained.keys():
return preTrained[modelv](weights='DEFAULT')
else:
logging.error(f'No model found for {modelv}')
return None
def train(model, optimizer, lossFn, trainLoader, valLoader, epochs, device, modelName):
best_model = 1000
for epoch in range(epochs):
trainingLoss = 0.0
trainingAcc = 0.0
totalt = 0
model.train()
for batch in tqdm(trainLoader, ncols= 70, desc= 'Training'):
optimizer.zero_grad()
inputs, targets = batch
inputs = inputs.to(device)
targets = targets.to(device)
output = model(inputs)
loss = lossFn(output, targets)
loss.backward()
optimizer.step()
trainingLoss += loss.item()
_, predicted = torch.max(output.data, 1)
trainingAcc += (predicted == targets).sum().item()
totalt += targets.size(0)
valLoss = 0.0
valAcc = 0.0
totalv = 0
model.eval()
with torch.no_grad():
for batch in tqdm(valLoader, ncols= 70, desc= 'Validating'):
inputs, targets = batch
inputs = inputs.to(device)
targets = targets.to(device)
output = model(inputs)
loss = lossFn(output, targets)
valLoss += loss.item()
_, predicted = torch.max(output.data, 1)
valAcc += (predicted == targets).sum().item()
totalv += targets.size(0)
if best_model > valAcc/totalv:
best_model = valAcc/totalv
torch.save(best_model, modelName + '_' + str(epoch) + '.pth')
print(f'Epoca: {epoch}, train_loss: {(trainingLoss/totalt):.4f}, train_acc: {(trainingAcc/totalt):.4f}')
print(f'Epoca: {epoch}, val_loss: {(valLoss/totalv):.4f}, val_acc: {(valAcc/totalv):.4f}')
print("------------------------------------")
def main():
args = get_args()
logging.basicConfig(level=logging.INFO, format='%(levelname)s: %(message)s')
device = checkingGPU()
logging.info(f'''Starting training:
Model: {args.model}
N_Dense: {args.num_cap}
Epochs: {args.epochs}
Batch size: {args.batch_size}
Learning rate: {args.lr}
Images dimension: {args.re_dim}
Device: {device}
''')
n_features= getNeurons(args.num_cap) if args.num_cap > 1 else [args.neurons]
trainData, valData = creatingData(args)
model = trainingModel(menuDisplay(args.model), output = args.clases,
n_layers= args.num_cap,
n_features= n_features)
model.to(device)
optimizer = optim.SGD(model.parameters(), lr= args.lr, momentum=0.9)
loss= torch.nn.CrossEntropyLoss()
try:
train(model, optimizer, loss, trainData, valData, args.epochs, device, args.model)
except torch.cuda.OutOfMemoryError:
logging.error('Detected OutOfMemoryError! '
'Enabling checkpointing to reduce memory usage, but this slows down training. '
'Consider enabling AMP (--amp) for fast and memory efficient training')
torch.cuda.empty_cache()
if __name__ == '__main__':
main()