rxn.reaction_preprocessing.stable_data_splitter.StableSplitter

class rxn.reaction_preprocessing.stable_data_splitter.StableSplitter(split_ratio, seed=0)[source]

Bases: object

Split data in a reproducible manner, based on the hash of values required to always be in the same split.

Useful for instance to ensure that a reaction product with a given SMILES will always be in the same split.

Parameters

  • split_ratio (float) –

  • seed (int, default: 0) –

__init__(split_ratio, seed=0)[source]
Parameters

  • split_ratio (float) – The approximate split ratio for test and validation set.

  • seed (int, default: 0) – seed to use for hashing. The default of 0 corresponds to the default value in the xxhash implementation.

Methods

__init__(split_ratio[, seed])

type split_ratio

float

get_split(split_value)

param split_value

Attributes

HASH_SIZE