coarse-grained-forecasting-benchmark's picture
Add anonymous reproduction code
41dd7e0 verified
Raw History Blame Contribute Delete
1.28 kB
"""The image metrics: csFID (the paper's metric), cFID and cKID.
backend.py the pinned Clean-FID pipeline: one resize (PIL bicubic, 299x299), the TF-Inception-v3 weights
(2015-12-05), the pool3 feature (cFID, cKID) and the spatial mixed_6/conv feature (csFID)
adapters.py the dataset adapter psi_D: physical state -> uint8 (H, W, 3) image, with constants fitted on
training data only (records/<system>/fid_adapter.json)
euler.py, passive_scalar.py, heterogeneous_waves(_augmented).py, nlkg.py the adapters of the five systems
evaluate.py the metrics of all test conditions of a model against the truth bank
"""
from .adapters import ADAPTERS, FIDAdapter, get_adapter, register_adapter # noqa: F401
from . import euler as _euler # noqa: F401 (registers euler_coarse_vorticity)
from . import passive_scalar as _passive_scalar # noqa: F401 (registers passive_scalar_coarse)
from . import heterogeneous_waves as _heterogeneous_waves # noqa: F401 (base class of the waves adapter)
from . import heterogeneous_waves_augmented as _heterogeneous_waves_augmented # noqa: F401 (registers heterogeneous_waves_augmented_coarse: waves)
from . import nlkg as _nlkg # noqa: F401 (registers nlkg_coarse_u: NLKG; reads the displacement u)