-
Notifications
You must be signed in to change notification settings - Fork 8
/
Prototypical.py
200 lines (162 loc) · 8.15 KB
/
Prototypical.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
"""
An implementation of the "Prototypical Networks for Few-shot Learning" in PyTorch,
trained and tested on bearing fault diagnosis probelm.
Data: 2022/11/13
Author: Xiaohan Chen
Email: [email protected]
"""
import argparse
import logging
import warnings
import os
import time
import numpy as np
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from Utils.logger import setlogger
from tqdm import *
from PrepareData.CWRU import CWRUloader
from PrototypicalNets import CNN1D
from Datasets import PrototypicalData
from Sampler import BatchSampler
from Loss.PrototypicalLoss import prototypical_loss as loss_fn
import Utils.utilis as utils
# ===== Define argments =====
def parse_args():
parser = argparse.ArgumentParser(description='Implementation of Prototypical Neural Networks')
# log files
parser.add_argument("--log_file", type=str, default="./logs/Prototypical.log", help="log file path")
# dataset information
parser.add_argument("--datadir", type=str, default="/home/xiaohan/codelab/datasets", help="data directory")
parser.add_argument("--s_load", type=int, default=3, help="source domain working condition")
parser.add_argument("--t_load", type=int, default=2, help="target domain working condition")
parser.add_argument("--s_label_set", type=list, default=[0,1,2,3,4,5,6,7,8,9], help="source domain label set")
parser.add_argument("--t_label_set", type=list, default=[0,1,2,3,4,5,6,7,8,9], help="target domain label set")
# pre-processing
parser.add_argument("--fft", type=bool, default=False, help="FFT preprocessing")
parser.add_argument("--window", type=int, default=128, help="time window, if not augment data, window=1024")
parser.add_argument("--normalization", type=str, default="0-1", choices=["None", "0-1", "mean-std"], help="normalization option")
# backbone
parser.add_argument("--backbone", type=str, default="CNN1D", choices=["ResNet1D", "MLPNet", "CNN1D"])
parser.add_argument("--savemodel", type=bool, default=False, help="whether save pre-trained model in the classification task")
parser.add_argument("--pretrained", type=bool, default=False, help="whether use pre-trained model in transfer learning tasks")
parser.add_argument("--n_train", type=int, default=500, help="The number of training data per class")
parser.add_argument("--n_val", type=int, default=200, help="the number of validation data per class")
parser.add_argument("--n_test", type=int, default=200, help="the number of test data per class")
parser.add_argument("--support", type=int, default=10, help="the number of support set per class")
parser.add_argument("--query", type=int, default=10, help="the number of query set per class")
parser.add_argument("--episodes", type=int, default=80, help="the number of episodes per epoch")
parser.add_argument("--showstep", type=int, default=50, help="show training history every 'showstep' steps")
# optimization & training
parser.add_argument("--num_workers", type=int, default=0, help="the number of dataloader workers")
parser.add_argument("--max_epoch", type=int, default=300)
parser.add_argument("--lr", type=float, default=1e-4, help="learning rate")
parser.add_argument("--optimizer", type=str, default="sgd", choices=["adam", "sgd"])
parser.add_argument('--gamma', type=float, default=0.8, help='learning rate scheduler parameter for step and exp')
args = parser.parse_args()
return args
# ===== Load Data =====
def loaddata(args):
source_data = CWRUloader(args, args.s_load, args.s_label_set, args.n_train)
target_data = CWRUloader(args, args.t_load, args.t_label_set, args.n_val+args.n_test)
val_data = {key:target_data[key][:args.n_val] for key in target_data.keys()}
test_data = {key:target_data[key][-args.n_test:] for key in target_data.keys()}
# convert the data format from dictionary to tensor
source_data = PrototypicalData.ProtitypicalData(source_data)
val_data = PrototypicalData.ProtitypicalData(val_data)
test_data = PrototypicalData.ProtitypicalData(test_data)
# source_data.y: all source data labels
source_sampler = BatchSampler.BatchSampler(source_data.y, args.support, args.query, args.episodes)
val_sampler = BatchSampler.BatchSampler(val_data.y, args.support, args.query, args.episodes)
test_sampler = BatchSampler.BatchSampler(test_data.y, args.support, args.query, episodes=args.episodes)
source_loader = DataLoader(source_data, batch_sampler=source_sampler)
val_loader = DataLoader(val_data, batch_sampler=val_sampler)
test_loader = DataLoader(test_data, batch_sampler=test_sampler)
return source_loader, val_loader, test_loader
# ===== Evaluate the model =====
def Test(args, Net, dataloader):
Net.eval()
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
correct_num, error_num = 0, 0
data_iter = iter(dataloader)
acc_his, loss_his = [], []
for x,y in data_iter:
x, y = x.to(device), y.to(device)
with torch.no_grad():
outputs = Net(x)
loss, acc = loss_fn(outputs, target=y, n_support=args.support)
acc_his.append(acc.item())
loss_his.append(loss.item())
avg_acc = np.mean(acc_his)
avg_loss = np.mean(loss_his)
return avg_acc, avg_loss
# ===== Train the model =====
def Train(args):
# Consider the gpu or cpu condition
if torch.cuda.is_available():
device = torch.device("cuda")
device_count = torch.cuda.device_count()
logging.info('using {} gpus'.format(device_count))
else:
warnings.warn("gpu is not available")
device = torch.device("cpu")
device_count = 1
logging.info('using {} cpu'.format(device_count))
# load datasets
source_loader, val_loader, test_loader = loaddata(args)
# load the Prototypical Network
Net = CNN1D.CNN1D()
# Define optimizer and learning rate decay
parameter_list = [{"params": Net.parameters(), "lr": args.lr}]
optimizer = utils.optimizer(args, parameter_list)
Net.to(device)
# train
best_acc = 0.0
meters = {"train_acc": [], "train_loss": [], "val_acc": [], "val_loss": []}
pre_prototypes = torch.randn(10, 64)
for epoch in range(args.max_epoch):
Net.train()
train_iter = iter(source_loader) # len(train_iter) = episodes
acc_his, loss_his = [], []
for x,y in tqdm(train_iter, ncols = 70, leave=False):
# move to GPU if available
x = x.to(device) # [class*(num_support + num_querry), sample_length]
y = y.to(device) # [class*(num_support + num_querry)]
outputs = Net(x) # [class*(num_support + num_querry), feature_dim]
loss, acc = loss_fn(outputs, target=y, n_support=args.support)
# clear previous gradients, compute gradients
optimizer.zero_grad()
loss.backward()
# performs updates using calculated gradients
optimizer.step()
# accuracy and loss of per episode
acc_his.append(acc.item())
loss_his.append(loss.item())
train_acc = np.mean(acc_his)
train_loss = np.mean(loss_his)
meters["train_acc"].append(train_acc)
meters["train_loss"].append(train_loss)
# validation
val_acc, val_loss = Test(args, Net, val_loader)
meters["val_acc"].append(val_acc)
meters["val_loss"].append(val_loss)
# print training history
if epoch % args.showstep == 0:
logging.info("Epoch: {:>3}/{}, train_loss: {:.4f}, val_loss: {:.4f}, train_acc: {:6.2f}%, val_acc: {:6.2f}%".format(
epoch+1, args.max_epoch, train_loss, val_loss, train_acc*100, val_acc*100))
logging.info("="*15+"Done!"+"="*15)
if __name__ == "__main__":
if not os.path.exists("./History"):
os.makedirs("./History")
if not os.path.exists("./checkpoints"):
os.makedirs("./checkpoints")
args = parse_args()
# set the logger
if not os.path.exists("./logs"):
os.makedirs("./logs")
setlogger(args.log_file)
# save the args
for k, v in args.__dict__.items():
logging.info("{}: {}".format(k, v))
Train(args)