Skip to content

Train

ExperimentParams

Bases: dict

Container for experiment parameters

Source code in src/trecover/utils/train.py 
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
class ExperimentParams(dict):
    """ Container for experiment parameters """

    def __init__(self, params: Union[argparse.Namespace, Dict, None] = None):
        super(ExperimentParams, self).__init__()

        if isinstance(params, argparse.Namespace):
            self.__dict__.update(vars(params))
        if isinstance(params, dict):
            self.__dict__.update(params)

    def __getitem__(self, key: Any) -> Any:
        return self.__dict__[key]

    def __repr__(self):
        return repr(self.__dict__)

    def __str__(self):
        return str(self.__dict__)

    def jsonify(self) -> Dict:
        """
        Simplify experiment parameters for further json serialization.

        Returns
        -------
            Experiment parameters as a dictionary with python built-in types values.

        """
        simplified = dict()
        simplified.update(self.__dict__)

        for key, value in self.__dict__.items():
            if isinstance(value, Path):
                simplified[key] = str(value)

        return simplified

    def update(self, *args: Any, **kwargs: Any) -> None:
        self.__dict__.update(*args, **kwargs)

jsonify 

jsonify()

Simplify experiment parameters for further json serialization.

Returns:

Type Description
Experiment parameters as a dictionary with python built-in types values.
Source code in src/trecover/utils/train.py 
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
def jsonify(self) -> Dict:
    """
    Simplify experiment parameters for further json serialization.

    Returns
    -------
        Experiment parameters as a dictionary with python built-in types values.

    """
    simplified = dict()
    simplified.update(self.__dict__)

    for key, value in self.__dict__.items():
        if isinstance(value, Path):
            simplified[key] = str(value)

    return simplified

set_seeds 

set_seeds(seed=2531)

Set seeds for experiment reproducibility

Source code in src/trecover/utils/train.py 
57
58
59
60
61
62
63
def set_seeds(seed: int = 2531) -> None:
    """ Set seeds for experiment reproducibility """

    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)  # multi-GPU

get_experiment_params 

get_experiment_params(parser, args=None)

Parse command line arguments as ExperimentParams object.

Parameters:

Name Type Description Default
parser ArgumentParser

Object for parsing command line strings into Python objects.

 required
args Optional[List[str]], default

Command line arguments.

 None

Returns:

Name Type Description
ExperimentParams ExperimentParams

Parsed command line arguments as ExperimentParams object.
Source code in src/trecover/utils/train.py 
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
def get_experiment_params(parser: ArgumentParser, args: Optional[List[str]] = None) -> ExperimentParams:
    """
    Parse command line arguments as ExperimentParams object.

    Parameters
    ----------
    parser : ArgumentParser
        Object for parsing command line strings into Python objects.
    args : Optional[List[str]], default=None
        Command line arguments.

    Returns
    -------
    ExperimentParams:
        Parsed command line arguments as ExperimentParams object.

    """

    return ExperimentParams(parser.parse_args(args=args))

load_params 

load_params(model_params)

Get experiment parameters container.

Parameters:

Name Type Description Default
model_params Path

Path to serialized experiment parameters.

 required

Returns:

Name Type Description
ExperimentParams ExperimentParams

Experiment parameters container as a ExperimentParams object.
Source code in src/trecover/utils/train.py 
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
def load_params(model_params: Path) -> ExperimentParams:
    """
    Get experiment parameters container.

    Parameters
    ----------
    model_params : Path
        Path to serialized experiment parameters.

    Returns
    -------
    ExperimentParams:
        Experiment parameters container as a ExperimentParams object.

    """

    return ExperimentParams(json.load(model_params.open()))

save_params 

save_params(data, filepath, sort=False)

Save experiment parameters on disk.

Parameters:

Name Type Description Default
data Dict

Experiment parameters.

 required
filepath Path

File path for saving.

 required
sort bool, default

Perform parameters keys sorting.

 False
Source code in src/trecover/utils/train.py 
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
def save_params(data: Dict, filepath: Path, sort=False) -> None:
    """
    Save experiment parameters on disk.

    Parameters
    ----------
    data : Dict
        Experiment parameters.
    filepath : Path
        File path for saving.
    sort : bool, default=False
        Perform parameters keys sorting.

    """

    with filepath.open('w') as f:
        json.dump(data, indent=2, fp=f, sort_keys=sort)

get_experiment_mark 

get_experiment_mark()

Generate a string mark based on current date

Source code in src/trecover/utils/train.py 
125
126
127
128
129
def get_experiment_mark() -> str:
    """ Generate a string mark based on current date """

    date = datetime.now()
    return f'{date.month:0>2}-{date.day:0>2}-{date.hour:0>2}-{date.minute:0>2}'

optimizer_to_str 

optimizer_to_str(optimizer)

Get optimizer object string representation.

Parameters:

Name Type Description Default
optimizer Optimizer

Optimizer object for representation.

 required

Returns:

Name Type Description
str str

Optimizer string representation.
Source code in src/trecover/utils/train.py 
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
def optimizer_to_str(optimizer: Optimizer) -> str:
    """
    Get optimizer object string representation.

    Parameters
    ----------
    optimizer : Optimizer
        Optimizer object for representation.

    Returns
    -------
    str:
        Optimizer string representation.

    """

    optimizer_str_repr = str(optimizer)

    optimizer_name = optimizer_str_repr.split()[0]
    optimizer_params = re.findall(r'(.+): (.+)', optimizer_str_repr)

    return f"{optimizer_name}({', '.join([f'{param.strip()}={value.strip()}' for param, value in optimizer_params])})"

transfer 

transfer(tensors, to_device)

Transfer the tensors to a specified device.

Parameters:

Name Type Description Default
tensors Tuple

Sequence of tensors to transfer.

 required
to_device torch.device

The desired device of returned tensors.

 required

Returns:

Type Description
Tuple[Optional[Tensor], ...]

Tuple of tensors allocated on the specified device.
Source code in src/trecover/utils/train.py 
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
def transfer(tensors: Tuple[Optional[Tensor], ...], to_device: torch.device) -> Tuple[Optional[Tensor], ...]:
    """
    Transfer the tensors to a specified device.

    Parameters
    ----------
    tensors : Tuple
        Sequence of tensors to transfer.
    to_device : torch.device
        The desired device of returned tensors.

    Returns
    -------
    Tuple[Optional[Tensor], ...]:
        Tuple of tensors allocated on the specified device.

    """
    return tuple([
        tensor.to(to_device) if isinstance(tensor, Tensor) else tensor
        for tensor in tensors
    ])