noether.data.datasets.cfd.simshift_heatsink.config

Classes

SimshiftHeatsinkConfig

Configuration for the SIMSHIFT Heatsink dataset.

Module Contents

class noether.data.datasets.cfd.simshift_heatsink.config.SimshiftHeatsinkConfig(/, **data)

Bases: noether.core.schemas.dataset.DatasetBaseConfig

Configuration for the SIMSHIFT Heatsink dataset.

This dataset uses HDF5 files from the SIMSHIFT benchmark for unsupervised domain adaptation of neural surrogates for physical simulations.

If root is not set, the dataset is automatically downloaded from the HuggingFace Hub (simshift/SIMSHIFT_data, heatsink.zip) and cached locally.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

Parameters:

data (Any)

kind: str | None = 'noether.data.datasets.cfd.SimshiftHeatsinkDataset'

Kind of dataset to use.

root: str | None = None

Root directory of the dataset. If None, auto-downloads from HuggingFace Hub.

split: Literal['train', 'val', 'test']

Which split of the dataset to use. Must be one of “train”, “val”, or “test”.

difficulty: Literal['easy', 'medium', 'hard'] | None = None

Domain-gap difficulty level between source and target domains. If None, load all difficulties.

domain: Literal['source', 'target'] | None = None

source (in-distribution) or target (shifted). If None, load both.

Type:

Which domain to load

splits_path: str | None = None

Path to the splits.json file. If None, defaults to {root}/splits.json.

metadata_path: str | None = None

Path to the metadata.csv file. If None, defaults to {root}/metadata.csv.