|
a |
|
b/survival4D/config.py |
|
|
1 |
import os |
|
|
2 |
import json |
|
|
3 |
from pathlib import Path |
|
|
4 |
from pyhocon import ConfigTree, ConfigFactory |
|
|
5 |
|
|
|
6 |
|
|
|
7 |
def get_conf(conf: ConfigTree, group: str = "", key: str = "", default=None): |
|
|
8 |
if group: |
|
|
9 |
key = ".".join([group, key]) |
|
|
10 |
return conf.get(key, default) |
|
|
11 |
|
|
|
12 |
|
|
|
13 |
class BaseConfig: |
|
|
14 |
GROUP = "" |
|
|
15 |
|
|
|
16 |
@classmethod |
|
|
17 |
def from_conf(cls, conf_path: Path): |
|
|
18 |
raise NotImplementedError("Must be implemented by subclasses") |
|
|
19 |
|
|
|
20 |
def save(self, output_dir: Path): |
|
|
21 |
assert output_dir.is_dir(), "output_dir has to be a directory." |
|
|
22 |
with open(str(output_dir.joinpath(self.__class__.__name__)) + ".json", "w") as file: |
|
|
23 |
json.dump(self.__dict__, file, indent=4) |
|
|
24 |
|
|
|
25 |
def to_dict(self): |
|
|
26 |
return self.__dict__ |
|
|
27 |
|
|
|
28 |
|
|
|
29 |
class ExperimentConfig(BaseConfig): |
|
|
30 |
GROUP = "experiment" |
|
|
31 |
|
|
|
32 |
def __init__(self, data_path: Path, output_dir: Path, n_evals: int, n_bootstraps: int, n_folds: int, |
|
|
33 |
search_method: str): |
|
|
34 |
self.data_path = data_path |
|
|
35 |
self.output_dir = output_dir |
|
|
36 |
self.n_evals = n_evals |
|
|
37 |
self.n_folds = n_folds |
|
|
38 |
self.n_bootstraps = n_bootstraps |
|
|
39 |
self.search_method = search_method |
|
|
40 |
|
|
|
41 |
|
|
|
42 |
class NNExperimentConfig(ExperimentConfig): |
|
|
43 |
def __init__( |
|
|
44 |
self, data_path: Path, output_dir: Path, n_evals: int, n_bootstraps: int, n_folds: int, search_method: str, |
|
|
45 |
batch_size: int, n_epochs: int, backend: str |
|
|
46 |
): |
|
|
47 |
super().__init__( |
|
|
48 |
data_path=data_path, output_dir=output_dir, n_evals=n_evals, n_bootstraps=n_bootstraps, n_folds=n_folds, |
|
|
49 |
search_method=search_method |
|
|
50 |
) |
|
|
51 |
self.batch_size = batch_size |
|
|
52 |
self.n_epochs = n_epochs |
|
|
53 |
self.backend = backend |
|
|
54 |
|
|
|
55 |
@classmethod |
|
|
56 |
def from_conf(cls, conf_path): |
|
|
57 |
conf = ConfigFactory.parse_file(str(conf_path)) |
|
|
58 |
data_path = Path(os.path.abspath(get_conf(conf, group=cls.GROUP, key="data_path"))) |
|
|
59 |
if get_conf(conf, group=cls.GROUP, key="output_dir") is None: |
|
|
60 |
output_dir = data_path.parent.joinpath("output") |
|
|
61 |
else: |
|
|
62 |
output_dir = Path(get_conf(conf, group=cls.GROUP, key="output_dir")) |
|
|
63 |
return cls( |
|
|
64 |
data_path=data_path, |
|
|
65 |
output_dir=output_dir, |
|
|
66 |
batch_size=get_conf(conf, group=cls.GROUP, key="batch_size", default=16), |
|
|
67 |
n_epochs=get_conf(conf, group=cls.GROUP, key="n_epochs", default=100), |
|
|
68 |
n_evals=get_conf(conf, group=cls.GROUP, key="n_evals", default=50), |
|
|
69 |
n_bootstraps=get_conf(conf, group=cls.GROUP, key="n_bootstraps", default=100), |
|
|
70 |
n_folds=get_conf(conf, group=cls.GROUP, key="n_folds", default=6), |
|
|
71 |
search_method=get_conf(conf, group=cls.GROUP, key="search_method", default="particle swarm"), |
|
|
72 |
backend=get_conf(conf, group=cls.GROUP, key="backend", default="torch") |
|
|
73 |
) |
|
|
74 |
|
|
|
75 |
|
|
|
76 |
class CoxExperimentConfig(ExperimentConfig): |
|
|
77 |
@classmethod |
|
|
78 |
def from_conf(cls, conf_path): |
|
|
79 |
conf = ConfigFactory.parse_file(str(conf_path)) |
|
|
80 |
data_path = Path(os.path.abspath(get_conf(conf, group=cls.GROUP, key="data_path"))) |
|
|
81 |
if get_conf(conf, group=cls.GROUP, key="output_dir") is None: |
|
|
82 |
output_dir = data_path.parent.joinpath("output") |
|
|
83 |
else: |
|
|
84 |
output_dir = Path(get_conf(conf, group=cls.GROUP, key="output_dir")) |
|
|
85 |
return cls( |
|
|
86 |
data_path=data_path, |
|
|
87 |
output_dir=output_dir, |
|
|
88 |
n_evals=get_conf(conf, group=cls.GROUP, key="n_evals", default=50), |
|
|
89 |
n_bootstraps=get_conf(conf, group=cls.GROUP, key="n_bootstraps", default=100), |
|
|
90 |
n_folds=get_conf(conf, group=cls.GROUP, key="n_folds", default=6), |
|
|
91 |
search_method=get_conf(conf, group=cls.GROUP, key="search_method", default="particle swarm") |
|
|
92 |
) |
|
|
93 |
|
|
|
94 |
|
|
|
95 |
class HypersearchConfig(BaseConfig): |
|
|
96 |
GROUP = "hypersearch" |
|
|
97 |
|
|
|
98 |
def __init__(self, **kwargs): |
|
|
99 |
for key in kwargs: |
|
|
100 |
setattr(self, key, kwargs[key]) |
|
|
101 |
|
|
|
102 |
@classmethod |
|
|
103 |
def from_conf(cls, conf_path: Path): |
|
|
104 |
conf = ConfigFactory.parse_file(str(conf_path)) |
|
|
105 |
conf = getattr(conf, cls.GROUP) |
|
|
106 |
return cls(**conf) |
|
|
107 |
|
|
|
108 |
|
|
|
109 |
class ModelConfig(BaseConfig): |
|
|
110 |
GROUP = "model" |
|
|
111 |
|
|
|
112 |
def __init__(self, **kwargs): |
|
|
113 |
for key in kwargs: |
|
|
114 |
setattr(self, key, kwargs[key]) |
|
|
115 |
|
|
|
116 |
@classmethod |
|
|
117 |
def from_conf(cls, conf_path: Path): |
|
|
118 |
conf = ConfigFactory.parse_file(str(conf_path)) |
|
|
119 |
conf = getattr(conf, cls.GROUP) |
|
|
120 |
return cls(**conf) |