-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathtrain.py
More file actions
215 lines (180 loc) · 9.33 KB
/
train.py
File metadata and controls
215 lines (180 loc) · 9.33 KB
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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
import torch
import os
from torch.utils.data import DataLoader, random_split, WeightedRandomSampler
from sklearn.model_selection import StratifiedShuffleSplit
from torch.utils.data import Subset
from torch.optim import Adam
import torch.nn as nn
from dataset import MouseDataset
from model import MouseModel, VisionModel, AudioModel, BehaviorModel
from tqdm import tqdm
import argparse
from collections import Counter
from torch.utils.tensorboard import SummaryWriter
from sklearn.metrics import accuracy_score
# import torch.distributed as dist
# from torch.nn.parallel import DistributedDataParallel
# from torch.utils.data.distributed import DistributedSampler
import numpy as np
def parse_args():
parser = argparse.ArgumentParser()
# data path
# parser.add_argument('--video_path', default="/data/zhaozhenghao/Projects/Mouse_behavior/dataset/Formalin/frame_folder", type=str, dest='video_path', help='Video path.')
parser.add_argument('--pred_path', default='alphapose-results.json', type=str, dest='pred_path', help='Prediction path.')
parser.add_argument('--label_path', default='Formalin_acute_pain_1.xlsx', type=str, dest='label_path', help='Label path.')
# parser.add_argument('--audio_path', default='/data/zhaozhenghao/Projects/Mouse_behavior/dataset/Formalin/Formalin_Ultrasound_recording.wav', type=str, dest='audio_path', help='Audio path.')
# hyperparameters
parser.add_argument('--learning_rate', type=float, default=0.0001)
parser.add_argument('--num_epochs', type=int, default=50)
parser.add_argument('--batch_size', type=int, default=32)
# parameters
parser.add_argument('--resampling_rate', type=int, default=1500, help='Resampling rate for audio.')
parser.add_argument('--step', type=int, default=10, help='Step size for sliding window.')
parser.add_argument('--stride', type=int, default=10, help='Stride size for sliding window.')
# tensorboard
parser.add_argument('--tensorboard', type=bool, default=True)
parser.add_argument('--log_dir', type=str, default='./logs')
# ddp
# parser.add_argument("--local_rank", type=int, default=0)
args = parser.parse_args()
return args
def uniform_dist(dataset):
labels = np.array([labels for _, labels in dataset])
sss = StratifiedShuffleSplit(n_splits=1, test_size=0.2, random_state=0)
labels_str = [f"{label_tensor[0].item()}{label_tensor[1].item()}"
for label_tensor in labels]
# Placeholder for X since we only need to stratify based on the labels
X_dummy = np.zeros(len(labels))
# Use 'labels_str' for stratification
for train_index, test_index in sss.split(X_dummy, labels_str):
train_dataset = Subset(dataset, train_index)
test_dataset = Subset(dataset, test_index)
return train_dataset, test_dataset
def get_flat_labels(dataset):
labels = []
for _, label_tensor in dataset:
labels.append(label_tensor[0].item() * 2 + label_tensor[1].item()) # This maps [0,0]->0, [0,1]->1, [1,0]->2, [1,1]->3
return labels
# Calculate sample weights
def calculate_weights(dataset):
flat_labels = get_flat_labels(dataset)
class_sample_count = torch.tensor([(torch.tensor(flat_labels) == t).sum() for t in torch.unique(torch.tensor(flat_labels), sorted=True)])
weight = 1. / class_sample_count.float()
sample_weights = torch.tensor([weight[t] for t in flat_labels])
print(sample_weights)
return sample_weights
def main(args):
# local_rank = args.local_rank
# dist.init_process_group(backend="nccl", init_method="env://")
local_rank = 0
# torch.cuda.set_device(local_rank)
# device = torch.device('cuda', local_rank)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
if args.tensorboard and local_rank == 0:
writer = SummaryWriter(args.log_dir)
# dataset
dataset = MouseDataset(args.pred_path, args.label_path)
#Maintaining the distribution in train and test set
train_dataset, test_dataset = uniform_dist(dataset)
#
train_weights = calculate_weights(train_dataset)
train_sampler = WeightedRandomSampler(train_weights, len(train_weights))
# Calculate weights for the testing dataset
test_weights = calculate_weights(test_dataset)
test_sampler = WeightedRandomSampler(test_weights, len(test_weights))
# Now, use the samplers in your DataLoader
train_loader = DataLoader(train_dataset, batch_size=32, sampler=train_sampler)
test_loader = DataLoader(test_dataset, batch_size=32, sampler=test_sampler)
# for steps, (behavior_feat, labels) in enumerate(tqdm(train_loader)):
# rnn_1_neuron =RNNVanilla(behavior_feat[0][0],1)
# train_loader = DataLoader(train_dataset, batch_size=args.batch_size, shuffle=False, sampler=train_sampler, num_workers=0)
# test_loader = DataLoader(test_dataset, batch_size=args.batch_size, shuffle=False, sampler=test_sampler, num_workers=0)
# for steps, (behavior_feat, labels) in enumerate(tqdm(train_loader)):
# print("behavior_feat : ", behavior_feat)
# print("behavior_feat_size :",behavior_feat.shape)
# print("behavior_feat[0] : ", behavior_feat[0])
# print("behavior_feat[0].size :",behavior_feat[0].shape)
# print("labels : ",labels)
# print("labels_size :",labels.shape)
# break
#'''
# model
# model = VisionModel(num_meta_features=12, num_classes=3).to(device)
# model = AudioModel(num_audio_features=50, num_classes=3).to(device)
# model = MouseModel(num_meta_features=12, num_audio_features=50, num_classes=3).to(device)
model = BehaviorModel(num_features=12, num_classes=2).to(device) #
# model = DistributedDataParallel(model, device_ids=[local_rank], output_device=local_rank, find_unused_parameters=True)
criterion = nn.BCEWithLogitsLoss()
#criterion = nn.BCELoss()
## criterion = nn.MSELoss()
# optimizer = Adam(model.parameters(), lr=args.learning_rate)
optimizer = torch.optim.Adagrad(model.parameters(), lr=0.01, lr_decay=0, weight_decay=0, initial_accumulator_value=0, eps=1e-10)
best_valid_loss = float('inf')
for epoch in range(args.num_epochs):
model.train()
all_preds = []
all_labels = []
total_loss = 0
for steps, (behavior_feat, labels) in enumerate(tqdm(train_loader)):
behavior_feat, labels = behavior_feat.to(device), labels.to(device)
# outputs = model(images, behavior_feat)
outputs = model(behavior_feat)
# outputs = model(audio)
# outputs = model(images, behavior_feat, audio)
# print(outputs, labels)
loss = criterion(outputs, labels.float())
# loss = creiterion(outputs, behavior_feat.float())
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_loss += loss.item()
probs = torch.sigmoid(outputs)
preds = (probs > 0.5).float() # Convert logits to binary predictions
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())
train_accuracy = accuracy_score(all_labels, all_preds)
train_loss = total_loss / len(train_loader)
if args.tensorboard and local_rank == 0:
writer.add_scalar('Train/Accuracy', train_accuracy, epoch)
writer.add_scalar('Train/Loss', train_loss, epoch)
print(f"Epoch [{epoch+1}/{args.num_epochs}], Train Loss: {train_loss:.4f}, Train Accuracy: {train_accuracy:.2f}")
model.eval()
total_loss = 0
all_preds = []
all_labels = []
with torch.no_grad():
for behavior_feat, labels in test_loader:
behavior_feat, labels = behavior_feat.to(device), labels.to(device)
# outputs = model(images, behavior_feat)
outputs = model(behavior_feat)
# outputs = model(audio)
# outputs = model(images, behavior_feat, audio)
loss = criterion(outputs, labels.float())
total_loss += loss
# probs = torch.sigmoid(outputs)
preds = (outputs > 0.5).float() # Convert logits to binary predictions
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())
test_loss = total_loss / len(test_loader)
val_accuracy = accuracy_score(all_labels, all_preds)
if test_loss < best_valid_loss:
best_valid_loss = test_loss
checkpoint_path = 'best_model_checkpoint.pt'
torch.save({
'epoch': epoch,
'model_state_dict': model.state_dict(),
'optimizer_state_dict': optimizer.state_dict(),
'loss': test_loss,
}, checkpoint_path)
if args.tensorboard and local_rank == 0:
writer.add_scalar('Test/Accuracy', val_accuracy, epoch)
writer.add_scalar('Test/Loss', test_loss, epoch)
print(f"Epoch [{epoch+1}/{args.num_epochs}], Testing Loss: {test_loss:.4f}, Test Accuracy: {val_accuracy:.2f}")
print("Training finished!")
if args.tensorboard:
writer.close()
#'''
if __name__ == "__main__":
args = parse_args()
main(args)
# torch.multiprocessing.spawn(main, args=(args,))