-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_adaptation.py
80 lines (61 loc) · 2.74 KB
/
run_adaptation.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
from argparse import ArgumentParser
import os
from pathlib import Path
import numpy as np
import torch
import yaml
from models import BaseNet
from utils.get_accuracy import get_accuracy
from utils.get_datamodule_cls import get_datamodule_cls
from utils.get_tta_cls import get_tta_cls
from utils.seed import seed_everything
CHECKPOINT_PATH = os.path.join(Path(__file__).resolve().parents[1], "checkpoints")
CONFIG_DIR = os.path.join(Path(__file__).resolve().parents[1], "configs")
DEFAULT_CONFIG = "tta_entropy_minimization.yaml"
def run_adaptation(config):
# load source config
with open(os.path.join(CHECKPOINT_PATH, config["source_run"], "config.yaml")) as f:
source_config = yaml.safe_load(f)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
datamodule_cls = get_datamodule_cls(source_config["dataset_name"])
model_cls = BaseNet
tta_cls = get_tta_cls(config["tta_method"])
if source_config["subject_ids"] == "all":
subject_ids = datamodule_cls.all_subject_ids
else:
subject_ids = [source_config["subject_ids"]]
source_config["preprocessing"]["alignment"] = False
source_config["preprocessing"]["batch_size"] = 1
datamodule = datamodule_cls(source_config["preprocessing"], subject_ids=subject_ids)
cal_accs, test_accs = [], []
for version, subject_id in enumerate(subject_ids):
seed_everything(source_config["seed"])
# load checkpoint
ckpt_path = os.path.join(CHECKPOINT_PATH, config["source_run"], str(subject_id),
"model-v1.ckpt")
model = model_cls.load_from_checkpoint(ckpt_path, map_location=device)
# set subject_id
datamodule.subject_id = subject_id
datamodule.prepare_data()
datamodule.setup()
model = tta_cls(model, config["tta_config"], datamodule.info)
if config.get("continual", False):
cal_acc = get_accuracy(model, datamodule.calibration_dataloader(), device)
cal_accs.append(cal_acc)
print(f"cal_acc subject {subject_id}: {100 * cal_accs[-1]:.2f}%")
acc = get_accuracy(model, datamodule.test_dataloader(), device)
test_accs.append(acc)
print(f"test_acc subject {subject_id}: {100 *test_accs[-1]:.2f}%")
# print overall test accuracy
if config.get("continual", False):
print(f"cal_acc: {100 * np.mean(cal_accs):.2f}")
print(f"test_acc: {100 * np.mean(test_accs):.2f}")
if __name__ == "__main__":
# parse arguments
parser = ArgumentParser()
parser.add_argument("--config", default=DEFAULT_CONFIG)
args = parser.parse_args()
# load config
with open(os.path.join(CONFIG_DIR, args.config)) as f:
config = yaml.safe_load(f)
run_adaptation(config)