-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_baselines.py
164 lines (135 loc) · 7.86 KB
/
run_baselines.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
import argparse
import os
import numpy as np
import pandas as pd
import pytorch_lightning as pl
import torch
from sklearn.metrics import roc_auc_score
from torchaudio.transforms import MelSpectrogram
from torchvision.transforms import ToTensor
from torchvision.transforms.v2 import Compose, Normalize
from baselines.classifiers import AudioClassifier, ImageClassifier, MultimodalClassifier, TextClassifier
from baselines.de_model import DEModel
from baselines.dirichlet import DirichletModel
from baselines.mc_model import MCDModel
from data_generation.text_processing import extract_deep_text_features
from dataset import LUMADataset
pl.seed_everything(42)
class Text2FeatureTransform():
def __init__(self, features_path):
with open(features_path, 'rb') as f:
self.features = np.load(f)
def __call__(self, text, idx):
return self.features[idx]
class PadCutToSizeAudioTransform():
def __init__(self, size):
self.size = size
def __call__(self, audio):
if audio.shape[-1] < self.size:
audio = torch.nn.functional.pad(audio, (0, self.size - audio.shape[-1]))
elif audio.shape[-1] > self.size:
audio = audio[:, :self.size]
return audio
def parse_args() -> tuple[argparse.Namespace, list[str]]:
parser = argparse.ArgumentParser()
parser.add_argument('-n', '--noise_type', type=str, default='')
args, unknown = parser.parse_known_args()
return args, unknown
args, unknown = parse_args()
if args.noise_type not in ['', 'diversity', 'label', 'sample']:
raise ValueError('Invalid noise type')
train_audio_data_path = test_audio_data_path = ood_audio_data_path = 'data/audio' if args.noise_type != 'sample' else 'data/noisy_audio'
suffix = ''
if args.noise_type == 'diversity':
suffix = '_diversity'
elif args.noise_type == 'label':
suffix = '_noise_label'
elif args.noise_type == 'sample':
suffix = '_noise'
train_audio_path = f'{train_audio_data_path}/datalist_train{suffix if args.noise_type != "sample" else ""}.csv'
test_audio_path = f'{test_audio_data_path}/datalist_test{suffix if args.noise_type not in ("sample", "diversity") else ""}.csv'
ood_audio_path = f'{ood_audio_data_path}/datalist_ood{suffix if args.noise_type != "sample" else ""}.csv'
train_image_path = f'data/image_data_train{suffix}.pickle'
train_text_path = f'data/text_data_train{suffix}.tsv'
test_image_path = f'data/image_data_test{suffix if args.noise_type != "diversity" else ""}.pickle'
test_text_path = f'data/text_data_test{suffix if args.noise_type != "diversity" else ""}.tsv'
ood_image_path = f'data/image_data_ood{suffix}.pickle'
ood_text_path = f'data/text_data_ood{suffix}.tsv'
extract_deep_text_features(train_text_path, output_path=f'text_features_train_{args.noise_type}.npy')
extract_deep_text_features(test_text_path, output_path=f'text_features_test_{args.noise_type}.npy')
extract_deep_text_features(ood_text_path, output_path=f'text_features_ood_{args.noise_type}.npy')
print(f'Loading data from {train_audio_path}, {test_audio_path}, {ood_audio_path}, {train_image_path}, {train_text_path}, {test_image_path}, {test_text_path}, {ood_image_path}, {ood_text_path}')
image_transform = Compose([
ToTensor(),
# Resize((224, 224)),
Normalize(mean=(0.51, 0.49, 0.44),
std=(0.27, 0.26, 0.28))
])
train_dataset = LUMADataset(train_image_path, train_audio_path, train_audio_data_path, train_text_path,
text_transform=Text2FeatureTransform(f'text_features_train_{args.noise_type}.npy'),
audio_transform=Compose([MelSpectrogram(), PadCutToSizeAudioTransform(128)]),
image_transform=image_transform)
test_dataset = LUMADataset(test_image_path, test_audio_path, test_audio_data_path, test_text_path,
text_transform=Text2FeatureTransform(f'text_features_test_{args.noise_type}.npy'),
audio_transform=Compose([MelSpectrogram(), PadCutToSizeAudioTransform(128)]),
image_transform=image_transform)
ood_dataset = LUMADataset(ood_image_path, ood_audio_path, ood_audio_data_path, ood_text_path,
text_transform=Text2FeatureTransform(f'text_features_ood_{args.noise_type}.npy'),
audio_transform=Compose([MelSpectrogram(), PadCutToSizeAudioTransform(128)]),
image_transform=image_transform, ood=True)
train_dataset, val_dataset = torch.utils.data.random_split(train_dataset, [int(0.8 * len(train_dataset)),
len(train_dataset) - int(
0.8 * len(train_dataset))])
batch_size = 128
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=8)
val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=8)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=batch_size, shuffle=False, num_workers=8)
ood_loader = torch.utils.data.DataLoader(ood_dataset, batch_size=batch_size, shuffle=False, num_workers=8)
# Now we can use the loaders to train a model
acc_dict = {}
classes = 42
mc_samples = 100
dropout_p = 0.3
n_ensemble = 10
mc_models = [MCDModel(c, classes, mc_samples, dropout_p) for c in [ImageClassifier, AudioClassifier, TextClassifier,
MultimodalClassifier]]
de_models = [DEModel(c, classes, n_ensemble, dropout_p) for c in [ImageClassifier, AudioClassifier, TextClassifier,
MultimodalClassifier]]
dir_models = [DirichletModel(MultimodalClassifier, classes, dropout=dropout_p)]
models = mc_models + de_models + dir_models
uncertainty_values = {}
for classifier in models:
model = classifier
try:
model_name = classifier.__class__.__name__ + '_' + classifier.model.__class__.__name__
except AttributeError:
model_name = classifier.__class__.__name__ + '_' + classifier.models[0].__class__.__name__
trainer = pl.Trainer(max_epochs=300,
gpus=1 if torch.cuda.is_available() else 0,
callbacks=[pl.callbacks.EarlyStopping(monitor='val_loss', patience=10, mode='min'),
pl.callbacks.ModelCheckpoint(monitor='val_loss', mode='min', save_last=True)])
trainer.fit(model, train_loader, val_loader)
print('Testing model')
trainer.test(model, test_loader)
acc_dict[model_name] = trainer.callback_metrics["test_acc"]
acc_dict[model_name + '_ale'] = trainer.callback_metrics["test_ale"]
acc_dict[model_name + '_entropy_ep'] = trainer.callback_metrics["test_entropy_epi"]
aleatoric_uncertainties = model.aleatoric_uncertainties
epistemic_uncertainties = model.epistemic_uncertainties
print('Testing OOD')
trainer.test(model, ood_loader)
acc_dict[model_name + '_ood_ale'] = trainer.callback_metrics["test_ale"]
acc_dict[model_name + '_ood'] = trainer.callback_metrics["test_acc"]
acc_dict[model_name + '_ood_entropy_ep'] = trainer.callback_metrics["test_entropy_epi"]
aleatoric_uncertainties_ood = model.aleatoric_uncertainties
epistemic_uncertainties_ood = model.epistemic_uncertainties
auc_score = roc_auc_score(
np.concatenate([np.zeros(len(epistemic_uncertainties)), np.ones(len(epistemic_uncertainties_ood))]),
np.concatenate([epistemic_uncertainties, epistemic_uncertainties_ood]))
uncertainty_values[f'{model_name}_epistemic'] = epistemic_uncertainties
uncertainty_values[f'{model_name}_aleatoric'] = aleatoric_uncertainties
acc_dict[model_name + '_ood_auc'] = auc_score
for key, value in acc_dict.items():
print(f'{key}: {value}')
uncertainty_df = pd.DataFrame(uncertainty_values)
uncertainty_df.to_csv(f'uncertainty_values{args.noise_type}.csv')