"""
Hyperparameter defaults and valid ranges for self-evolving training.
"""
from dataclasses import dataclass
@dataclass
class TuneConstant:
"""Hyperparameter defaults and validation bounds.
Attributes:
default_example_num: Number of examples per iteration
default_iteration_num: Default training iterations
default_max_sampled_example_num: Max examples to sample
default_parallel_num: Default parallelism for inference
default_max_num_sample_error_cases: Max error cases to log
default_early_stop_score: Score threshold for early stopping
min_iteration_num: Minimum iterations allowed
max_iteration_num: Maximum iterations allowed
min_parallel_num: Minimum parallelism allowed
max_parallel_num: Maximum parallelism allowed
min_example_num: Minimum examples allowed
max_example_num: Maximum examples allowed
"""
default_example_num: int = 1
default_iteration_num: int = 3
default_max_sampled_example_num: int = 10
default_parallel_num: int = 1
default_max_num_sample_error_cases: int = 10
default_early_stop_score: float = 1.0
min_iteration_num: int = 1
max_iteration_num: int = 20
min_parallel_num: int = 1
max_parallel_num: int = 20
min_example_num: int = 0
max_example_num: int = 20