"""Hyper-Parameters of a simulated task based on variants of the Forrester function:
.. math:: f(x) = ((\alpha x - 2)^2) sin(\beta x - 4)
This task uses a meta-model that is trained using a dataset of points from different functions, each
with different values of alpha and beta. This meta-model can then be used to sample "fake" points
from a given forrester function.
"""
import typing
from dataclasses import dataclass
from typing import Callable, ClassVar, Dict, List, Tuple
from orion.benchmark.task.profet.model_utils import get_architecture_forrester
from orion.benchmark.task.profet.profet_task import ProfetTask
if typing.TYPE_CHECKING:
import torch
[docs]class ProfetForresterTask(ProfetTask):
"""Simulated Task consisting in training a model on a variant of the Forrester function."""
[docs] @dataclass
class ModelConfig(ProfetTask.ModelConfig):
"""Config for training the Profet model on a Forrester task."""
benchmark: str = "forrester"
# ---------- "Abstract" class attributes:
json_file_name: ClassVar[str] = "data_sobol_forrester.json"
get_architecture: ClassVar[
Callable[[int], "torch.nn.Module"]
] = get_architecture_forrester
""" Callable that takes the input dimensionality and returns the network to be trained. """
hidden_space: ClassVar[int] = 2
normalize_targets: ClassVar[bool] = True
log_cost: ClassVar[bool] = False
log_target: ClassVar[bool] = False
shapes: ClassVar[Tuple[Tuple[int, ...], Tuple[int, ...], Tuple[int, ...]]] = (
(10, 1),
(9, 10),
(9, 10),
)
y_min: ClassVar[float] = -18.049155413936802
y_max: ClassVar[float] = 14718.31848526001
c_min: ClassVar[float] = -18.049155413936802
c_max: ClassVar[float] = 14718.3184852600
# -----------
[docs] def call(self, x: float) -> List[Dict]:
return super().call(x=x)
[docs] def get_search_space(self) -> Dict[str, str]:
return {
"x": "uniform(0.0, 1.0, discrete=False)",
}