|
1
|
|
|
import logging |
|
|
|
|
|
|
2
|
|
|
import os |
|
3
|
|
|
from typing import Dict, List, Union |
|
4
|
|
|
|
|
5
|
|
|
import yaml |
|
6
|
|
|
|
|
7
|
|
|
from deepreg.config.v011 import parse_v011 |
|
8
|
|
|
|
|
9
|
|
|
|
|
10
|
|
|
def update_nested_dict(d: Dict, u: Dict) -> Dict: |
|
11
|
|
|
""" |
|
12
|
|
|
Merge two dicts. |
|
13
|
|
|
|
|
14
|
|
|
https://stackoverflow.com/questions/3232943/update-value-of-a-nested-dictionary-of-varying-depth |
|
15
|
|
|
|
|
16
|
|
|
:param d: dict to be overwritten in case of conflicts. |
|
17
|
|
|
:param u: dict to be merged into d. |
|
18
|
|
|
:return: |
|
19
|
|
|
""" |
|
20
|
|
|
|
|
21
|
|
|
for k, v in u.items(): |
|
22
|
|
|
if isinstance(v, dict): |
|
23
|
|
|
d[k] = update_nested_dict(d.get(k, {}), v) |
|
24
|
|
|
else: |
|
25
|
|
|
d[k] = v |
|
26
|
|
|
return d |
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
|
def load_configs(config_path: Union[str, List[str]]) -> Dict: |
|
30
|
|
|
""" |
|
31
|
|
|
Load multiple configs and update the nested dictionary. |
|
32
|
|
|
|
|
33
|
|
|
:param config_path: list of paths or one path. |
|
34
|
|
|
:return: the loaded config |
|
35
|
|
|
""" |
|
36
|
|
|
if isinstance(config_path, str): |
|
37
|
|
|
config_path = [config_path] |
|
38
|
|
|
# replace ~ with user home path |
|
39
|
|
|
config_path = [os.path.expanduser(x) for x in config_path] |
|
40
|
|
|
config: Dict = {} |
|
41
|
|
|
for config_path_i in config_path: |
|
42
|
|
|
with open(config_path_i) as file: |
|
43
|
|
|
config_i = yaml.load(file, Loader=yaml.FullLoader) |
|
44
|
|
|
config = update_nested_dict(d=config, u=config_i) |
|
45
|
|
|
loaded_config = config_sanity_check(config) |
|
46
|
|
|
|
|
47
|
|
|
if loaded_config != config: |
|
48
|
|
|
# config got updated |
|
49
|
|
|
head, tail = os.path.split(config_path[0]) |
|
50
|
|
|
filename = "updated_" + tail |
|
51
|
|
|
save(config=loaded_config, out_dir=head, filename=filename) |
|
52
|
|
|
logging.error( |
|
|
|
|
|
|
53
|
|
|
f"Used config is outdated. " |
|
54
|
|
|
f"An updated version has been saved at " |
|
55
|
|
|
f"{os.path.join(head, filename)}" |
|
56
|
|
|
) |
|
57
|
|
|
|
|
58
|
|
|
return loaded_config |
|
59
|
|
|
|
|
60
|
|
|
|
|
61
|
|
|
def save(config: dict, out_dir: str, filename: str = "config.yaml"): |
|
62
|
|
|
""" |
|
63
|
|
|
Save the config into a yaml file. |
|
64
|
|
|
|
|
65
|
|
|
:param config: configuration to be outputed |
|
66
|
|
|
:param out_dir: directory of the output file |
|
67
|
|
|
:param filename: name of the output file |
|
68
|
|
|
""" |
|
69
|
|
|
assert filename.endswith(".yaml") |
|
70
|
|
|
with open(os.path.join(out_dir, filename), "w+") as f: |
|
71
|
|
|
f.write(yaml.dump(config)) |
|
72
|
|
|
|
|
73
|
|
|
|
|
74
|
|
|
def config_sanity_check(config: dict) -> dict: |
|
|
|
|
|
|
75
|
|
|
""" |
|
76
|
|
|
Check if the given config satisfies the requirements. |
|
77
|
|
|
|
|
78
|
|
|
:param config: entire config. |
|
79
|
|
|
""" |
|
80
|
|
|
|
|
81
|
|
|
# check data |
|
82
|
|
|
data_config = config["dataset"] |
|
83
|
|
|
|
|
84
|
|
|
if data_config["type"] not in ["paired", "unpaired", "grouped"]: |
|
85
|
|
|
raise ValueError(f"data type must be paired / unpaired / grouped, got {type}.") |
|
86
|
|
|
|
|
87
|
|
|
if data_config["format"] not in ["nifti", "h5"]: |
|
88
|
|
|
raise ValueError(f"data format must be nifti / h5, got {format}.") |
|
89
|
|
|
|
|
90
|
|
|
assert "dir" in data_config |
|
91
|
|
|
for mode in ["train", "valid", "test"]: |
|
92
|
|
|
assert mode in data_config["dir"].keys() |
|
93
|
|
|
data_dir = data_config["dir"][mode] |
|
94
|
|
|
if data_dir is None: |
|
95
|
|
|
logging.warning(f"Data directory for {mode} is not defined.") |
|
|
|
|
|
|
96
|
|
|
if not (isinstance(data_dir, (str, list)) or data_dir is None): |
|
97
|
|
|
raise ValueError( |
|
98
|
|
|
f"data_dir for mode {mode} must be string or list of strings," |
|
99
|
|
|
f"got {data_dir}." |
|
100
|
|
|
) |
|
101
|
|
|
|
|
102
|
|
|
# back compatibility support |
|
103
|
|
|
config = parse_v011(config) |
|
104
|
|
|
|
|
105
|
|
|
# check model |
|
106
|
|
|
if config["train"]["method"] == "conditional": |
|
107
|
|
|
if data_config["labeled"] is False: # unlabeled |
|
108
|
|
|
raise ValueError( |
|
109
|
|
|
"For conditional model, data have to be labeled, got unlabeled data." |
|
110
|
|
|
) |
|
111
|
|
|
|
|
112
|
|
|
return config |
|
113
|
|
|
|
DeepRegNet /
DeepReg
Loading history...