GaPFlow.Database

GaPFlow.Database#

class GaPFlow.Database(md: Any, db: dict, num_extra_features: int = 1)#

Bases: object

Container for GP training datasets.

Handles dataset initialization, normalization, data addition, and optional dtool integration for persistent dataset storage.

Parameters:

md (GaPFlow.md.MolecularDynamics) – An instance of the MD runner object. Adding a data point will lead to calling its run method.
db (dict) –
Configuration dictionary with keys:
- 'dtool_path' : str, path where training data is stored and loaded from.
- 'init_size' : int, minimum dataset size.
- 'init_width' : float, relative sampling width.
- 'init_method' : str, name of the (quasi-)random initialization method (‘lhc’, ‘rand’, ‘sobol’).
- 'init_seed' : int, random seed for initialization.
num_extra_features (int, number of additional features (next to solution, gap height + gradients)) – stored with the database (default is 1)

__init__(md: Any, db: dict, num_extra_features: int = 1) → None#

Methods

`__init__`(md, db[, num_extra_features])
`add_data`(Xnew)	Add new data entries to the database.
`get_readme_list_local`()	Get list of dtool README files for existing MD runs from a local directory.
`get_readme_list_remote`()	Get list of dtool README files for existing MD runs from a remote data server (via dtool_lookup_api)
`initialize`(Xtest[, dim])	Initialize database.
`set_training_path`(new_path[, check_temporary])	Set training path.
`write`()	Write the dataset arrays to disk (if the simulation output path is specified).

Attributes

`X_scale`	Normalization constants for input features.
`X_shift`	Normalization constants for input features.
`Xtrain`	Normalized input features of shape (Ntrain, Nfeat).
`Xtrain_target`	Normalized input features of shape (Ntrain, Nfeat).
`Y_scale`	Normalization constants for observations
`Y_shift`	Normalization constants for observations
`Ytrain`	Normalized observations of shape (Ntrain, 13).
`Ytrain_err`	Normalized observation error of shape (Ntrain, 13).
`config`	Configuration parameters of the database object.
`has_mock_md`	Flag that indicates whether the attached MD runner is a 'mock' object.
`md_config`	Configuration parameters of the attached MD runner object.
`num_features`	Number of possible features, actual ones are selected from GP's active_dims.
`output_path`	Simulation output path
`size`	Number of training samples currently stored.
`training_path`	Local storage location of dtool datasets.

property X_scale: Float[Array, 'Ntrain Nfeat']#: Normalization constants for input features.

property X_shift: Float[Array, 'Ntrain Nfeat']#: Normalization constants for input features.

property Xtrain: Float[Array, 'Ntrain Nfeat']#: Normalized input features of shape (Ntrain, Nfeat).

property Xtrain_target: Float[Array, 'Ntrain Nfeat']#: Normalized input features of shape (Ntrain, Nfeat).

property Y_scale: Float[Array, 'Ntrain 13']#: Normalization constants for observations

property Y_shift: Float[Array, 'Ntrain 13']#: Normalization constants for observations

property Ytrain: Float[Array, 'Ntrain 13']#: Normalized observations of shape (Ntrain, 13).

property Ytrain_err: Float[Array, 'Ntrain 13']#: Normalized observation error of shape (Ntrain, 13).

add_data(Xnew: Float[Array, 'Ntrain Nfeat']) → None#

Add new data entries to the database.

Parameters:: Xnew (jax.Array) – New samples of shape (Nnew, Nfeat).

property config: dict#: Configuration parameters of the database object.

get_readme_list_local()#

Get list of dtool README files for existing MD runs from a local directory.

Returns:: List of dicts containing the readme content
Return type:: list

get_readme_list_remote()#

Get list of dtool README files for existing MD runs from a remote data server (via dtool_lookup_api)

In the future, one should be able to pass a valid MongoDB query string to select data.

Returns:: List of dicts containing the readme content
Return type:: list

property has_mock_md: bool#: Flag that indicates whether the attached MD runner is a ‘mock’ object.

initialize(Xtest: Float[Array, 'Ntrain Nfeat'], dim: int = 1) → Float[Array, 'Ntrain Nfeat']#

Initialize database.

Parameters:

Xtest (jax.Array) – Candidate test points of shape (n_test, 6).
dim (int) – Dimension of the fluid problem (either 1 or 2, defaults to 1)

property md_config: dict#: Configuration parameters of the attached MD runner object.

property num_features: int#: Number of possible features, actual ones are selected from GP’s active_dims.

property output_path: str#: Simulation output path

set_training_path(new_path: str, check_temporary: bool = False) → None#

Set training path.

This modifies the storage location of dtool basepaths, also for the attached MD runner object.

Parameters:: new_path (str) – Training path

property size: int#: Number of training samples currently stored.

property training_path: str#: Local storage location of dtool datasets.

write() → None#: Write the dataset arrays to disk (if the simulation output path is specified).

GaPFlow.Database

Contents

GaPFlow.Database#