Skip to content

schema

Defines the schema for configuration validation using Sparkwheel's validation with dataclasses.

AdapterConfig dataclass

Adapter configuration for a specific mode.

Source code in src/lighter/engine/schema.py 
 9
10
11
12
13
14
15
16
@dataclass
class AdapterConfig:
    """Adapter configuration for a specific mode."""

    batch: Optional[dict] = None
    criterion: Optional[dict] = None
    metrics: Optional[dict] = None
    logging: Optional[dict] = None

AdaptersConfig dataclass

Adapters configuration for all modes.

Source code in src/lighter/engine/schema.py 
27
28
29
30
31
32
33
34
@dataclass
class AdaptersConfig:
    """Adapters configuration for all modes."""

    train: Optional[dict] = None  # Can be AdapterConfig but keep flexible
    val: Optional[dict] = None
    test: Optional[dict] = None
    predict: Optional[dict] = None

ArgsConfig dataclass

Arguments to pass to Trainer stage methods.

Source code in src/lighter/engine/schema.py 
70
71
72
73
74
75
76
77
@dataclass
class ArgsConfig:
    """Arguments to pass to Trainer stage methods."""

    fit: Optional[dict] = None
    validate: Optional[dict] = None
    test: Optional[dict] = None
    predict: Optional[dict] = None

DataloadersConfig dataclass

Dataloaders configuration for different stages.

Source code in src/lighter/engine/schema.py 
46
47
48
49
50
51
52
53
@dataclass
class DataloadersConfig:
    """Dataloaders configuration for different stages."""

    train: Optional[dict] = None
    val: Optional[dict] = None
    test: Optional[dict] = None
    predict: Optional[dict] = None

LighterConfig dataclass

Main Lighter configuration schema.

Source code in src/lighter/engine/schema.py 
80
81
82
83
84
85
86
87
88
@dataclass
class LighterConfig:
    """Main Lighter configuration schema."""

    trainer: dict  # pytorch_lightning.Trainer
    system: SystemConfig  # lighter.System
    project: Optional[str] = None
    vars: Optional[dict] = None
    args: Optional[ArgsConfig] = None

MetricsConfig dataclass

Metrics configuration for different stages.

Source code in src/lighter/engine/schema.py 
37
38
39
40
41
42
43
@dataclass
class MetricsConfig:
    """Metrics configuration for different stages."""

    train: Optional[list | dict] = None
    val: Optional[list | dict] = None
    test: Optional[list | dict] = None

PredictAdapterConfig dataclass

Adapter configuration for predict mode (no criterion).

Source code in src/lighter/engine/schema.py 
19
20
21
22
23
24
@dataclass
class PredictAdapterConfig:
    """Adapter configuration for predict mode (no criterion)."""

    batch: Optional[dict] = None
    logging: Optional[dict] = None

SystemConfig dataclass

System configuration with model, optimizer, scheduler, etc.

Source code in src/lighter/engine/schema.py 
56
57
58
59
60
61
62
63
64
65
66
67
@dataclass
class SystemConfig:
    """System configuration with model, optimizer, scheduler, etc."""

    model: Optional[dict] = None
    criterion: Optional[dict] = None
    optimizer: Optional[dict] = None
    scheduler: Optional[dict] = None
    inferer: Optional[dict] = None
    metrics: Optional[MetricsConfig] = None
    dataloaders: Optional[DataloadersConfig] = None
    adapters: Optional[AdaptersConfig] = None