Config API¶
AnnealConfig
¶
Bases: BaseModel
Configuration for curriculum annealing schedules.
Config
¶
Bases: BaseModel
Main configuration for NAICS training.
CurriculumConfig
¶
Bases: BaseModel
Structure-Aware Dynamic Curriculum (SADC) scheduler configuration.
validate_phase_boundaries()
¶
Ensure curriculum phases progress monotonically.
DataConfig
¶
Bases: BaseModel
Data configuration.
DataLoaderConfig
¶
Bases: BaseModel
Data loading and preprocessing configuration.
warn_large_batch(v)
classmethod
¶
Warn about potentially problematic batch sizes.
DirConfig
¶
Bases: BaseModel
File system directory configuration.
DistancesConfig
¶
Bases: BaseModel
Configuration for computing pairwise distances.
from_yaml(yaml_path)
classmethod
¶
Load configuration from YAML file.
DownloadConfig
¶
Bases: BaseModel
Configuration for downloading and preprocessing NAICS data.
from_yaml(yaml_path)
classmethod
¶
Load configuration from YAML file.
FalseNegativeConfig
¶
Bases: BaseModel
Configuration for handling false negatives during training.
GraphConfig
¶
Bases: BaseModel
Base configuration for HGCN training.
from_yaml(yaml_path)
classmethod
¶
Load GraphConfig from YAML file.
LoRAConfig
¶
Bases: BaseModel
LoRA (Low-Rank Adaptation) configuration.
LossConfig
¶
Bases: BaseModel
Loss function configuration.
MoEConfig
¶
Bases: BaseModel
Mixture of Experts configuration.
validate_top_k_vs_experts()
¶
Ensure top_k doesn't exceed num_experts.
ModelConfig
¶
Bases: BaseModel
Model architecture configuration.
RelationsConfig
¶
Bases: BaseModel
Configuration for computing pairwise relations.
from_yaml(yaml_path)
classmethod
¶
Load configuration from YAML file.
SamplingConfig
¶
Bases: BaseModel
Top-level sampling configuration (data layer strategies).
SansStaticConfig
¶
Bases: BaseModel
Configuration for static SANS-style sampling buckets.
StreamingConfig
¶
Bases: BaseModel
Configuration for streaming dataset.
validate_difficulty_ratios()
¶
Ensure easy + semi <= 1.0 at both start and end (hard is derived).
TokenizationConfig
¶
Bases: BaseModel
Configuration for tokenization caching.
from_yaml(yaml_path)
classmethod
¶
Load configuration from YAML file.
TrainerConfig
¶
TrainingConfig
¶
Bases: BaseModel
Optimizer and training configuration.
TripletsConfig
¶
Bases: BaseModel
Configuration for generating training triplets.
from_yaml(yaml_path)
classmethod
¶
Load configuration from YAML file.
load_config(config_class, yaml_path)
¶
Generic configuration loader for any Pydantic model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
config_class
|
Type[T]
|
The Pydantic model class to instantiate |
required |
yaml_path
|
Union[str, Path]
|
Path to YAML config file (absolute, relative, or under conf/) |
required |
Returns:
| Type | Description |
|---|---|
T
|
Instance of config_class with values from YAML |
Example
cfg = load_config(DownloadConfig, 'data/download.yaml')
parse_override_value(value)
¶
Parse override value from string to appropriate type.