chemicalchecker.util.splitter.ae_siam_traintest.AE_SiameseTraintest

class AE_SiameseTraintest(hdf5_file, split, replace_nan=None)

Bases: object

AE_SiameseTraintest class.

Initialize a AE_SiameseTraintest instance.

We assume the file is containing diffrent splits. e.g. “x_train”, “y_train”, “x_test”, …

Methods

close	Close the HDF5.
generator_fn	Return the generator function that we can query for batches.
get_split_indeces	Get random indeces for different splits.
get_split_names	Return the name of the splits.
get_sw	Get a batch of X.
get_x	Get a batch of X.
get_x_shapes	Return the shpaes of X.
get_xy	Get a batch of X and Y.
get_xy_shapes	Return the shpaes of X an Y.
open	Open the HDF5.
split_h5_blocks	Create the HDF5 file with validation splits from an input file.

Attributes

sw_name_right

available_splits = self.get_split_names() if split not in available_splits: raise Exception("Split '%s' not found in %s!" % (split, str(available_splits)))

close()

Close the HDF5.

static generator_fn(file_name, split, batch_size=None, only_x=False, sample_weights=False, shuffle=True, return_on_epoch=False)

Return the generator function that we can query for batches.

static get_split_indeces(rows, fractions)

Get random indeces for different splits.

get_split_names()

Return the name of the splits.

get_sw(beg_idx, end_idx)

Get a batch of X.

get_x(beg_idx, end_idx)

Get a batch of X.

get_x_shapes()

Return the shpaes of X.

get_xy(beg_idx, end_idx, shuffle)

Get a batch of X and Y.

get_xy_shapes()

Return the shpaes of X an Y.

open()

Open the HDF5.

static split_h5_blocks(in_file, out_file, split_names=['train', 'test', 'validation'], split_fractions=[0.8, 0.1, 0.1], block_size=1000, input_datasets=None)

Create the HDF5 file with validation splits from an input file.

Parameters:

• in_file (str) – path of the h5 file to read from.

• out_file (str) – path of the h5 file to write.

• split_names (list(str)) – names for the split of data.

• split_fractions (list(float)) – fraction of data in each split.

• block_size (int) – size of the block to be used.

• dataset (list) – only split the given dataset and ignore others.

sw_name_right

available_splits = self.get_split_names() if split not in available_splits:

raise Exception(“Split ‘%s’ not found in %s!” %

(split, str(available_splits)))