Basic molecular weight splitter class

    Splits internal compounds into train/validation/test using the MW calculated

    by SMILES string.

    """

    np.testing.assert_almost_equal(frac_train + frac_valid + frac_test, 1.)

    np.random.seed(seed)

    smiles_df = samples.compounds_df['smiles']

    mw = []

    for idx, row in smiles_df.iterrows():

        smiles = Chem.Mol

    # create new array that contains IDs sorted by MW

    df = samples.compounds_df

    mws = []

    for _, row in smiles_df.iterrows():

        mol = Chem.MolFromSmiles(row['smiles'])

        mw = Chem.rdMolDescriptors.CalcExactMolWt(mol)

        mws.append(mw)

    # split based on frac_train/frac_valid+frac_test

    # Sort by increasing MW

    mws = np.array(mws)

    sortidx = np.argsort(mws)

    # random split frac_valid/frac_test

    train_cutoff = frac_train * len(sortidx)

    valid_cutoff = (frac_train+frac_valid) * len(sortidx)

    train_cutoff = frac_train * len(samples.compounds_df)

    valid_cutoff = (frac_train+frac_valid) * len(samples.compounds_df)

    shuffled = np.random.permutation(range(len(samples.compounds_df)))

    # log stats on MW in each dataset

    return (shuffled[:train_cutoff], shuffled[train_cutoff:valid_cutoff],

            shuffled[valid_cutoff:])

    return (sortidx[:train_cutoff], sortidx[train_cutoff:valid_cutoff],

            sortidx[valid_cutoff:])

class RandomSplitter(Splitter):

  """