File size: 5,108 Bytes
07423df
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
import dataclasses
import logging
import os
from typing import Any, Dict, List, Optional

import numpy as np
from sqlitedict import SqliteDict

__all__ = ["Loggers"]

from llm_studio.src.utils.plot_utils import PLOT_ENCODINGS

logger = logging.getLogger(__name__)


def get_cfg(cfg: Any) -> Dict:
    """Returns simplified config elements

    Args:
        cfg: configuration

    Returns:
        Dict of config elements
    """

    items: Dict = {}
    type_annotations = cfg.get_annotations()

    cfg_dict = cfg.__dict__

    cfg_dict = {key: cfg_dict[key] for key in cfg._get_order(warn_if_unset=False)}

    for k, v in cfg_dict.items():
        if k.startswith("_") or cfg._get_visibility(k) < 0:
            continue

        if any([x in k for x in ["api"]]):
            continue

        if dataclasses.is_dataclass(v):
            elements_group = get_cfg(cfg=v)
            t = elements_group
            items = {**items, **t}
        else:
            type_annotation = type_annotations[k]
            if type_annotation == float:
                items[k] = float(v)
            else:
                items[k] = v

    return items


class NeptuneLogger:
    def __init__(self, cfg: Any):
        import neptune as neptune
        from neptune.utils import stringify_unsupported

        if cfg.logging._neptune_debug:
            mode = "debug"
        else:
            mode = "async"

        self.logger = neptune.init_run(
            project=cfg.logging.neptune_project,
            api_token=os.getenv("NEPTUNE_API_TOKEN", ""),
            name=cfg.experiment_name,
            mode=mode,
            capture_stdout=False,
            capture_stderr=False,
            source_files=[],
        )

        self.logger["cfg"] = stringify_unsupported(get_cfg(cfg))

    def log(self, subset: str, name: str, value: Any, step: Optional[int] = None):
        name = f"{subset}/{name}"
        self.logger[name].append(value, step=step)


class LocalLogger:
    def __init__(self, cfg: Any):
        logging.getLogger("sqlitedict").setLevel(logging.ERROR)

        self.logs = f"{cfg.output_directory}/charts.db"

        params = get_cfg(cfg)

        with SqliteDict(self.logs) as logs:
            logs["cfg"] = params
            logs.commit()

    def log(self, subset: str, name: str, value: Any, step: Optional[int] = None):
        if subset in PLOT_ENCODINGS:
            with SqliteDict(self.logs) as logs:
                if subset not in logs:
                    subset_dict = dict()
                else:
                    subset_dict = logs[subset]
                subset_dict[name] = value
                logs[subset] = subset_dict
                logs.commit()
            return

        # https://github.com/h2oai/wave/issues/447
        if np.isnan(value):
            value = None
        else:
            value = float(value)
        with SqliteDict(self.logs) as logs:
            if subset not in logs:
                subset_dict = dict()
            else:
                subset_dict = logs[subset]
            if name not in subset_dict:
                subset_dict[name] = {"steps": [], "values": []}

            subset_dict[name]["steps"].append(step)
            subset_dict[name]["values"].append(value)

            logs[subset] = subset_dict
            logs.commit()


class DummyLogger:
    def __init__(self, cfg: Optional[Any] = None):
        return

    def log(self, subset: str, name: str, value: Any, step: Optional[int] = None):
        return


class MainLogger:
    """Main logger"""

    def __init__(self, cfg: Any):
        self.loggers = {
            "local": LocalLogger(cfg),
            "external": Loggers.get(cfg.logging.logger),
        }

        try:
            self.loggers["external"] = self.loggers["external"](cfg)
        except Exception as e:
            logger.warning(
                f"Error when initializing logger. "
                f"Disabling custom logging functionality. "
                f"Please ensure logger configuration is correct and "
                f"you have a stable Internet connection: {e}"
            )
            self.loggers["external"] = DummyLogger(cfg)

    def reset_external(self):
        self.loggers["external"] = DummyLogger()

    def log(self, subset: str, name: str, value: str | float, step: float = None):
        for k, logger in self.loggers.items():
            if "validation_predictions" in name and k == "external":
                continue
            if subset == "internal" and not isinstance(logger, LocalLogger):
                continue
            logger.log(subset=subset, name=name, value=value, step=step)


class Loggers:
    """Loggers factory."""

    _loggers = {"None": DummyLogger, "Neptune": NeptuneLogger}

    @classmethod
    def names(cls) -> List[str]:
        return sorted(cls._loggers.keys())

    @classmethod
    def get(cls, name: str) -> Any:
        """Access to Loggers.

        Args:
            name: loggers name
        Returns:
            A class to build the Loggers
        """

        return cls._loggers.get(name, DummyLogger)