Simplified Molecular transformer example

applications/FLASK/Transformer/README.md

@@ -0,0 +1,10 @@
+# Molecular Transformer Baseline
+
+This is an encoder-decoder transformer on SMILES data.
+
+## Preparing the data
+
+The data can be prepared ahead of time to simplify the pipeline
+
+---
+This example is a simplified encoder-decoder transformer in the `application/nlp` directory. 

applications/FLASK/Transformer/arg_utils.py

@@ -0,0 +1,92 @@
+import argparse
+import dataset_utils
+
+
+# Command-line arguments
+def add_transformer_architecture_arguments(args: argparse.Namespace):
+    """
+    Adds the command line arguments to specify transformer architecture model
+    parameters. This is only relevant for the encoder-decoder transformer model.
+    """
+    args.add_argument(
+        "--num-attention-heads",
+        action="store",
+        default=8,
+        type=int,
+        help="number of parallel attention layers (default: 8)",
+        metavar="NUM",
+    )
+    args.add_argument(
+        "--embed-dim",
+        action="store",
+        default=512,
+        type=int,
+        help="embedding space dimension (default: 512)",
+        metavar="NUM",
+    )
+    args.add_argument(
+        "--feedforward-dim",
+        action="store",
+        default=0,
+        type=int,
+        help="feedforward network dimension. If zero, set to be "
+        "4 times the embedding dimension (default: 0)",
+        metavar="NUM",
+    )
+    args.add_argument(
+        "--num-layers",
+        action="store",
+        default=6,
+        type=int,
+        help="Number of encoder and decoder layers (default: 6)",
+        metavar="NUM",
+    )
+
+
+def add_dataset_arguments(args: argparse.Namespace, default: str):
+    """
+    Adds dataset-related arguments to an existing argparse object.
+    """
+    args.add_argument(
+        "--dataset",
+        type=str,
+        default=default,
+        help=f"Which dataset to use (default: {default})",
+        choices=dataset_utils.available_datasets(),
+    )
+    args.add_argument(
+        "--dataset-fraction",
+        action="store",
+        default=1.0,
+        type=float,
+        help="Fraction of dataset to use (default: 1.0)",
+        metavar="NUM",
+    )
+
+
+def add_training_arguments(parser: argparse.ArgumentParser):
+    parser.add_argument(
+        "--skip-validation",
+        action="store_true",
+        default=False,
+        help="Do not run validation (default: false)",
+    )
+    parser.add_argument(
+        "--always-shuffle",
+        action="store_true",
+        default=False,
+        help="Always shuffle training dataset, even if pretokenized (default: false)",
+    )
+    parser.add_argument(
+        "--validation-set-fraction",
+        type=float,
+        default=0.01,
+        help="Fraction of the validation dataset to use (default: 0.001)",
+    )
+    parser.add_argument(
+        "--save-prototext",
+        action="store_true",
+        default=False,
+        help="Save prototext experiment file instead of protobin (slower but "
+        "debuggable) (default: false)",
+    )

applications/FLASK/Transformer/dataset_utils.py

@@ -0,0 +1,26 @@
+import argparse
+import importlib
+import os
+import sys
+from typing import List
+
+dataset_dir = os.path.join(os.path.dirname(__file__), "datasets")
+
+
+def available_datasets() -> List[str]:
+    """
+    Returns the available datasets in the dataset folder.
+    """
+    result = []
+    for file in os.listdir(dataset_dir):
+        if file.endswith(".py"):
+            result.append(os.path.basename(file)[:-3])
+    return result
+
+
+def load_dataset(name: str):
+    """
+    Loads a dataset by importing the requested module.
+    """
+    sys.path.append(dataset_dir)
+    return importlib.import_module(name)

applications/FLASK/Transformer/datasets/QM9.py

@@ -0,0 +1,65 @@
+"""
+The QM9 dataset, stored as pre-tokenized binary files for optimized processing.
+"""
+
+import os
+import os.path
+
+import numpy as np
+from pretokenize.SMILES_tokenizer import MolTokenizer
+from pretokenize.data_utils import random_zero_array
+
+sequence_length = int(os.getenv("QM9_SEQUENCE_LENGTH", default="32"))
+
+# ----------------------------------------------
+# Setup
+# ----------------------------------------------
+
+# Load the datasets
+data_dir = os.getenv("QM9_DATA_DIR", "/p/vast1/lbann/datasets/FLASK/QM9")
+
+tokenizer = MolTokenizer(os.path.join(data_dir, "QM9_vocab.json"))
+tokenizer.load_vocab_file()
+
+dataset_train = np.load(os.path.join(data_dir, "QM9_Pretokenized.npy"))
+
+# dataset_train = np.zeros((140000, 32), dtype=np.float32)
+_vocab_size = 46
+pad_index = tokenizer.token_to_id("<pad>")
+sep_index = tokenizer.token_to_id("<eos>")
+
+# ----------------------------------------------
+# Sample access functions
+# ----------------------------------------------
+
+
+def num_train_samples():
+    return dataset_train.shape[0]
+
+
+def get_train_sample(i):
+    data = dataset_train[i]
+
+    boundary = np.where(data == sep_index)[0][0]
+    masked_data = random_zero_array(
+        data[:boundary], 0.15, tokenizer.token_to_id(tokenizer.mask_token)
+    )
+    output = np.zeros((2 * sequence_length), dtype=np.int32)
+    output[0:boundary] = masked_data
+    output[boundary] = sep_index
+    output[sequence_length:] = data
+    return output
+
+
+def sample_dims():
+    return (2 * sequence_length + 1,)
+
+
+def vocab_size():
+    return _vocab_size
+
+
+if __name__ == "__main__":
+    print("Training samples:", num_train_samples())
+    print("Training sample 101:")
+    print(get_train_sample(0))

applications/FLASK/Transformer/datasets/pretokenize/GenerateVocab.py

@@ -0,0 +1,100 @@
+from multiprocessing import Pool
+from SMILES_tokenizer import MolTokenizer
+from glob import glob
+import argparse
+import os
+import numpy as np
+from tqdm import tqdm
+
+
+parser = argparse.ArgumentParser(
+    description="Generate vocab files for different datasets"
+)
+
+parser.add_argument(
+    "--qm9", action="store_true", help="Generate vocab file for QM9 dataset"
+)
+parser.add_argument(
+    "--zinc", action="store_true", help="Generate vocab file for ZINC dataset"
+)
+parser.add_argument(
+    "--pubchem", action="store_true", help="Generate vocab file for PubChem dataset"
+)
+
+args = parser.parse_args()
+
+
+def join_vocabs(list_of_vocab_dicts):
+    """
+    Given a list of vocab dictionaries, join them together
+    such that all unique tokens are present in the final vocab dictionary
+    """
+    final_vocab = {}
+    counter = 0
+    for vocab in list_of_vocab_dicts:
+        for token in vocab.keys():
+            if token not in final_vocab.keys():
+                final_vocab[token] = counter
+                counter += 1
+    return final_vocab
+
+
+def generate_zinc_vocab_dict(smi_file):
+    tokenizer = MolTokenizer()
+    with open(smi_file, "r") as f:
+        data = f.readlines()
+    for i in tqdm(range(1, len(data))):
+        line = data[i].split(" ")
+        _ = tokenizer._tokenize(line[0])
+    return tokenizer.vocab_dict
+
+
+def main():
+
+    if args.qm9:
+        print("Generating vocab file for QM9 dataset")
+        tokenizer = MolTokenizer("QM9_vocab.json")
+        default_file = "/p/vast1/lbann/datasets/FLASK/QM9/QM9_smiles.txt"
+        qm9_file = os.getenv("QM9_FILE", default_file)
+        with open(qm9_file, "r") as smiles_data:
+            smiles_data = smiles_data.readlines()
+            for line in tqdm(smiles_data):
+                tokens = tokenizer.tokenize(line)
+        tokenizer.generate_vocab_file("QM9_vocab.json")
+        print("QM9 vocab file generated")
+
+    if args.zinc:
+        print("Generating vocab file for ZINC dataset")
+        default_dir = "/p/vast1/lbann/datasets/FLASK/ZINC"
+        zinc_dir = os.getenv("ZINC_DIR", default_dir)
+        zinc_files = glob(f"{zinc_dir}/*.smi")
+
+        print(len(zinc_files))
+
+        with Pool(20) as p:
+            zinc_vocab_dicts = p.map(generate_zinc_vocab_dict, zinc_files)
+
+        final_vocab = join_vocabs(zinc_vocab_dicts)
+
+        final_tokenizer = MolTokenizer("ZINC_vocab.json")
+        final_tokenizer.load_vocab_dict(final_vocab)
+        final_tokenizer.generate_vocab_file("ZINC_SMILES_vocab.json")
+        print("ZINC vocab file generated")
+
+    if args.pubchem:
+        print("Generating vocab file for PubChem dataset")
+        default_file = "/p/vast1/lbann/datasets/FLASK/pubchem/CID_SMILES_CANONICAL.smi"
+        pubchem_file = os.getenv("PUBCHEM_FILE", default_file)
+        with open(pubchem_file, "r") as smiles_data:
+            smiles_data = smiles_data.readlines()
+            tokenizer = MolTokenizer("PubChem_SMILES_vocab.json")
+            for line in tqdm(smiles_data):
+                smiles = line.split(" ")[1]
+                tokens = tokenizer.tokenize(smiles)
+
+            tokenizer.generate_vocab_file("PubChem_vocab.json")
+            print("PubChem vocab file generated")
+
+
+if __name__ == "__main__":
+    main()

applications/FLASK/Transformer/datasets/pretokenize/PubChem_vocab.json

@@ -0,0 +1 @@
+{"<pad>": 0, "<eos>": 1, "<bos>": 2, "<mask>": 3, "C": 4, "(": 5, "=": 6, "O": 7, ")": 8, "[O-]": 9, "[N+]": 10, "1": 11, "N": 12, "P": 13, "c": 14, "Cl": 15, "n": 16, "2": 17, "3": 18, "4": 19, "S": 20, "#": 21, "[PH]": 22, "[nH]": 23, "[N-]": 24, "[NH4+]": 25, "[As]": 26, "-": 27, "[Br-]": 28, "Br": 29, "[Ca+2]": 30, "[C-]": 31, "[O+]": 32, "5": 33, "[Cl-]": 34, ".": 35, "[Co]": 36, "o": 37, "[n+]": 38, "[S+]": 39, "s": 40, "[SeH2]": 41, "[Hg]": 42, "[NH3+]": 43, "[HH]": 44, "I": 45, "[K+]": 46, "[SeH]": 47, "[CH3-]": 48, "[Mg+2]": 49, "[Na+]": 50, "[Ni+2]": 51, "[Ni]": 52, "[OH-]": 53, "[H+]": 54, "[Se]": 55, "[Fe+3]": 56, "[W]": 57, "[NH+]": 58, "B": 59, "[NH2+]": 60, "F": 61, "[OH+]": 62, "6": 63, "[B-]": 64, "[S-]": 65, "[Si]": 66, "[Te]": 67, "7": 68, "[Al]": 69, "8": 70, "9": 71, "%10": 72, "%11": 73, "[nH+]": 74, "[Cd]": 75, "[Pt]": 76, "[Pt+2]": 77, "[NH2-]": 78, "[Co+2]": 79, "[NH-]": 80, "[Rh+2]": 81, "[I+]": 82, "[se]": 83, "[n-]": 84, "[Ni+3]": 85, "[Fe+2]": 86, "[Li+]": 87, "[Be+2]": 88, "[F-]": 89, "[Mn+3]": 90, "[I-]": 91, "[Pr]": 92, "[Ag+]": 93, "[CH-]": 94, "[Tc+6]": 95, "[Kr]": 96, "[Hg+]": 97, "[Zn]": 98, "[SnH]": 99, "[Au+]": 100, "[Al+3]": 101, "[SiH]": 102, "[Pb]": 103, "[Sn]": 104, "[SnH2]": 105, "[AlH2]": 106, "[CH2-]": 107, "[c-]": 108, "[cH-]": 109, "[P+]": 110, "[Zn+2]": 111, "[Sr+2]": 112, "[Cu+2]": 113, "[Ce]": 114, "[Cd+2]": 115, "[Cu]": 116, "[Bi]": 117, "[BiH3]": 118, "[SiH2]": 119, "[Fe]": 120, "[Pb+2]": 121, "[Sn+2]": 122, "[SbH3]": 123, "[P-]": 124, "[Si+]": 125, "[Ce+3]": 126, "[Ba+2]": 127, "[Cs+]": 128, "[AlH4-]": 129, "[U]": 130, "[Cu+]": 131, "[Ti+4]": 132, "[Sb]": 133, "[AlH]": 134, "[Tl+]": 135, "[Hg+2]": 136, "[Rb+]": 137, "[Hg-2]": 138, "[AsH2]": 139, "[Ge]": 140, "[Mn+2]": 141, "[H]": 142, "[AsH]": 143, "[SiH3]": 144, "[Bi+3]": 145, "[Cr+2]": 146, "[La+3]": 147, "[Cr+3]": 148, "[As+]": 149, "[PbH]": 150, "[AlH3]": 151, "[C]": 152, "[Al-]": 153, "[Ga]": 154, "[Be]": 155, "[Ca]": 156, "[P-3]": 157, "[Cr]": 158, "[Mg]": 159, "[O-2]": 160, "[Sb+3]": 161, "[SH-]": 162, "[Mn]": 163, "[Mo]": 164, "[S-2]": 165, "[Th]": 166, "[V]": 167, "[U+2]": 168, "p": 169, "[Sr]": 170, "[As+3]": 171, "[Sn+]": 172, "[SnH3]": 173, "[Cl+3]": 174, "[GeH]": 175, "[Zr]": 176, "[Cl+2]": 177, "[In+3]": 178, "[TeH2]": 179, "[Nd+3]": 180, "[As-]": 181, "[Dy]": 182, "[Es]": 183, "[Ir]": 184, "[La]": 185, "[PbH2]": 186, "[Lu]": 187, "[Np]": 188, "[Nd]": 189, "[Ne]": 190, "[Nb]": 191, "[Os]": 192, "[Pd]": 193, "[Pu]": 194, "[PoH2]": 195, "[Md]": 196, "[Pm]": 197, "[Pa]": 198, "[Re]": 199, "[Rh]": 200, "[Ru]": 201, "[Sm]": 202, "[Sc]": 203, "[SiH4]": 204, "[Ag]": 205, "[Ta]": 206, "[Tc]": 207, "[Tb]": 208, "[TlH]": 209, "[Tm]": 210, "[Ti]": 211, "[Ac]": 212, "[Am]": 213, "[Ar]": 214, "[AsH3]": 215, "[Bk]": 216, "[Cm]": 217, "[Er]": 218, "[Eu]": 219, "[Gd]": 220, "[GaH3]": 221, "[GeH4]": 222, "[Au]": 223, "[Hf]": 224, "[He]": 225, "[Ho]": 226, "[Xe]": 227, "[Yb]": 228, "[Y]": 229, "[AtH]": 230, "[Cf]": 231, "[Fm]": 232, "[InH3]": 233, "[Zr+4]": 234, "[I+2]": 235, "[Se+]": 236, "[Br+2]": 237, "[Cl+]": 238, "[Al-3]": 239, "[Tl]": 240, "[CH+]": 241, "[CH]": 242, "[H-]": 243, "%12": 244, "%13": 245, "%14": 246, "%15": 247, "%16": 248, "%17": 249, "%18": 250, "%19": 251, "%20": 252, "%21": 253, "%22": 254, "%23": 255, "%24": 256, "%25": 257, "%26": 258, "%27": 259, "%28": 260, "%29": 261, "%30": 262, "%31": 263, "%32": 264, "%33": 265, "%34": 266, "%35": 267, "%36": 268, "%37": 269, "%38": 270, "[In]": 271, "[No]": 272, "[Tb+3]": 273, "[Rn]": 274, "[Lu+3]": 275, "[Pd+2]": 276, "[Eu+3]": 277, "[Dy+3]": 278, "[Ho+3]": 279, "[Au+3]": 280, "[Sm+3]": 281, "[Y+3]": 282, "[Fe-3]": 283, "[Th+4]": 284, "[Au-]": 285, "[Sn-]": 286, "[Fe-]": 287, "[Tl+3]": 288, "[Sn+3]": 289, "[BH4-]": 290, "[Ta-2]": 291, "[Ir-2]": 292, "[Cr+6]": 293, "[o+]": 294, "[Te+4]": 295, "[Lr]": 296, "[N]": 297, "[V+2]": 298, "[Ga+3]": 299, "[Pt+4]": 300, "[Si-]": 301, "[Gd+3]": 302, "[Tc+3]": 303, "[Co+3]": 304, "[Pd-2]": 305, "[Pt-2]": 306, "[Sn+4]": 307, "[Er+3]": 308, "[Pd+]": 309, "[Se-2]": 310, "[Yb+3]": 311, "[Zn-2]": 312, "[O]": 313, "[Ta+5]": 314, "[Ba]": 315, "[Sc+3]": 316, "[SbH]": 317, "[Se+4]": 318, "[PH2-]": 319, "[Si+4]": 320, "[I+3]": 321, "[Ru+2]": 322, "[Ce+4]": 323, "[GeH2]": 324, "[Tm+3]": 325, "[Sb+]": 326, "[Se-]": 327, "[Ti+]": 328, "[Os+2]": 329, "[Ir+3]": 330, "[Ru+3]": 331, "[Pr+3]": 332, "[Na]": 333, "[K]": 334, "[Li]": 335, "[Rh+3]": 336, "[InH]": 337, "[Cs]": 338, "[Rb]": 339, "[Yb+2]": 340, "[Pt+]": 341, "[Ti+2]": 342, "[Mo+2]": 343, "[Ti+3]": 344, "[OH2+]": 345, "[Zr+2]": 346, "[C+]": 347, "[IrH+2]": 348, "[Mo+4]": 349, "[Eu+2]": 350, "[AlH2+]": 351, "[As+5]": 352, "[Sb-]": 353, "[Se+6]": 354, "[Cr+5]": 355, "[PH2]": 356, "[V+4]": 357, "[Ru+5]": 358, "[Li-]": 359, "[s+]": 360, "[AlH+2]": 361, "[c]": 362, "[BH3-]": 363, "[CH2]": 364, "[SnH4]": 365, "[S]": 366, "[PbH4]": 367, "[TeH]": 368, "[NH2]": 369, "[OH3+]": 370, "[At]": 371, "[te]": 372, "[siH]": 373, "[ReH]": 374, "[WH2]": 375, "[MoH2]": 376, "[GeH3]": 377, "[NH]": 378, "[B]": 379, "[AgH]": 380, "[TlH3]": 381, "[BH2]": 382, "[SH2+]": 383, "[Hf+4]": 384, "[Nb+5]": 385, "[FeH2]": 386, "[Cr+4]": 387, "[TaH2]": 388, "[GaH2]": 389, "[GaH]": 390, "[Th+2]": 391, "[P]": 392, "[SH]": 393, "[IH]": 394, "[si]": 395, "[Rh-]": 396, "[SH+]": 397, "[Rh+]": 398, "[Ru-]": 399, "[Rh-2]": 400, "[cH+]": 401, "[Tc+4]": 402, "[OH]": 403, "[Ru+]": 404, "[InH2]": 405, "[Zn-]": 406, "[Nb+2]": 407, "[Ru+4]": 408, "[Os+4]": 409, "[pH]": 410, "[Br+]": 411, "[PH4+]": 412, "[SiH3-]": 413, "[Zr-]": 414, "[Cu-2]": 415, "[Ta+2]": 416, "[Zr+3]": 417, "[Ni-]": 418, "[Al-2]": 419, "[Ru-2]": 420, "[CH2+]": 421, "[Ag-]": 422, "[BiH]": 423, "[Hg-]": 424, "[Tc+7]": 425, "[Os+8]": 426, "[Br+3]": 427, "[SbH2]": 428, "[Mo+3]": 429, "[BiH2]": 430, "[b-]": 431, "[Zn+]": 432, "[Fe+]": 433, "[Ru+6]": 434, "[BH-]": 435, "[Ni+]": 436, "[BH2-]": 437, "[Co+]": 438, "[PtH]": 439, "[Os+6]": 440, "[Te+]": 441, "[Bi+2]": 442, "[RuH]": 443, "[ZrH]": 444, "[SiH-]": 445, "[Ru+8]": 446, "[IH2+]": 447, "[ClH+]": 448, "[ClH2+]": 449, "[Os+7]": 450, "[SH3+]": 451, "[Os+]": 452, "[Tc+]": 453, "[Tc+5]": 454, "[IH+]": 455, "[PH+]": 456, "[Ru-3]": 457, "[W+2]": 458, "[PH2+]": 459, "[FeH]": 460, "[CoH2]": 461, "[W+]": 462, "[Fe+6]": 463, "[NiH]": 464, "[OsH]": 465, "[CH3+]": 466, "[p-]": 467, "[se+]": 468, "[PH-]": 469, "[I]": 470, "[Br]": 471, "[PH3+]": 472, "[CH3]": 473, "[Cu-]": 474, "[PbH2+2]": 475, "[Mn+]": 476, "[Ir+2]": 477, "[sH+]": 478, "[Cd-2]": 479, "[CuH]": 480, "[Re-]": 481, "[Re-2]": 482, "[TlH2]": 483, "[SeH2+]": 484, "[Ga-]": 485, "[Ir+]": 486, "[RhH]": 487, "[Gd+2]": 488, "[ZrH2]": 489, "[Os+5]": 490, "[Ni-2]": 491, "[Rh-3]": 492, "[RhH+2]": 493, "[Hf+2]": 494, "[Mn-2]": 495, "[Cr+]": 496, "[c+]": 497, "[Co-2]": 498, "[Nb-2]": 499, "[Fe-2]": 500, "[OsH2]": 501, "[PtH2]": 502, "[WH]": 503, "[RhH2]": 504, "[Ti-2]": 505, "[Bi+]": 506, "[SeH+]": 507, "[Al+2]": 508, "[SeH-]": 509, "[Nb+3]": 510, "[TaH]": 511, "[In-]": 512, "[Pt-]": 513, "[Tc+2]": 514, "[Nb-]": 515, "[Mo+]": 516, "[Ta-]": 517, "[Al+]": 518, "[NbH2]": 519, "[Pd-]": 520, "[FeH6-4]": 521, "[MoH]": 522, "[W-]": 523, "[In+]": 524, "[SbH+]": 525, "[AsH3+]": 526, "[Cl]": 527, "[F]": 528, "[PbH3]": 529, "[SnH3-]": 530, "[SeH3+]": 531, "[SeH3]": 532, "[TeH3+]": 533, "[TeH3]": 534, "[Ni-3]": 535, "[SbH4+]": 536, "[SbH4]": 537, "[AsH4+]": 538, "[PH-2]": 539, "[AsH3-]": 540, "[AsH2+]": 541, "[Cr-3]": 542, "[Cr-2]": 543, "[Fe-4]": 544, "[Mo-2]": 545, "[SiH2-]": 546, "[SiH2+]": 547, "[SnH2+]": 548, "[BiH2+2]": 549, "[Os-]": 550, "[PtH+]": 551, "[SbH5]": 552, "[p+]": 553, "[Po]": 554, "[Ra]": 555, "[Fr]": 556, "[AlH-]": 557, "[AlH2-]": 558, "[RuH+2]": 559, "[SnH+3]": 560, "[BiH2+]": 561, "[AsH-]": 562, "[RuH2+2]": 563, "[BH]": 564, "[Os-2]": 565, "[SiH3+]": 566, "[SnH3+]": 567, "[TeH2+]": 568, "[AsH5]": 569, "[TeH+]": 570, "[Si+3]": 571, "b": 572, "[Cr-]": 573, "[si+2]": 574, "[Si+2]": 575, "[Zr+]": 576, "[Co-]": 577, "[Ir-3]": 578, "[Mg+]": 579, "[V+]": 580, "[Bi-]": 581, "[Tb+4]": 582, "[P-2]": 583, "[SnH+2]": 584, "[Fe+5]": 585, "[Fe+4]": 586, "[BiH5]": 587, "[RuH+]": 588, "[RuH2]": 589, "[AlH6-3]": 590, "[IrH3]": 591, "[IrH4]": 592, "[ReH2]": 593, "[NbH3]": 594, "[RuH3]": 595, "[ReH3]": 596, "[IrH]": 597, "[IrH2]": 598, "[ZnH2]": 599, "[CrH2]": 600, "[LaH3]": 601, "[SmH3]": 602, "[EuH3]": 603, "[OsH-]": 604, "[WH4]": 605, "[MoH5]": 606, "[si+]": 607, "[RuH4]": 608, "[TaH3]": 609, "[MoH3]": 610, "[FeH4]": 611, "[MoH4]": 612, "[HgH2]": 613, "[ThH2]": 614, "[TiH2]": 615, "[AuH]": 616, "[ZrH3]": 617, "[si-]": 618, "[ReH7]": 619, "[UH2]": 620, "[OsH4]": 621, "[GaH4-]": 622, "[ThH4]": 623, "[WH6]": 624, "[oH+]": 625, "[RuH6]": 626, "[PdH2]": 627, "[PtH4]": 628, "[NiH2]": 629, "[VH2]": 630, "[PtH3]": 631, "[RhH3]": 632, "[UH3]": 633, "[MnH2]": 634, "[CuH2]": 635, "[CoH3]": 636, "[FeH3]": 637, "[TiH4]": 638, "[RuH5]": 639, "[bH-]": 640, "[BiH4]": 641, "[ReH4]": 642, "[HgH]": 643, "[WH3]": 644, "[CeH3]": 645, "[SnH2+2]": 646, "[Si-2]": 647, "%39": 648, "%40": 649, "%41": 650, "%42": 651, "%43": 652, "%44": 653, "%45": 654, "%46": 655, "%47": 656, "%48": 657, "%49": 658, "%50": 659, "%51": 660, "%52": 661, "%53": 662, "%54": 663, "%55": 664, "%56": 665, "%57": 666, "%58": 667, "%59": 668, "%60": 669, "%61": 670, "%62": 671, "[Hf+]": 672, "[al+]": 673, "[SnH+]": 674, "[Bi-2]": 675, "[AlH3-]": 676, "[PtH2+2]": 677, "[Zr-2]": 678, "[ClH4+3]": 679, "[pH+]": 680, "[Co-3]": 681, "%63": 682, "%64": 683, "%65": 684, "%66": 685, "%67": 686, "%68": 687, "%69": 688, "%70": 689, "%71": 690, "%72": 691, "%73": 692, "%74": 693, "[Pr+]": 694, "[te+]": 695, "[Os-3]": 696, "[al]": 697, "[PdH+]": 698, "[Ti-]": 699, "[SH2]": 700, "[BiH4-]": 701, "[TeH4]": 702, "[Zr-4]": 703, "[Zr-3]": 704, "[Ta+]": 705, "[U+4]": 706, "[U+3]": 707, "[ZrH2+2]": 708, "[Bh]": 709, "[Hs]": 710, "[Rf]": 711, "[Mt]": 712, "[Sg]": 713, "[Db]": 714, "[AsH4]": 715, "[UH]": 716, "[YH2]": 717, "[YH]": 718, "%75": 719, "%76": 720, "%77": 721, "%78": 722, "%79": 723, "%80": 724, "%81": 725, "%82": 726, "%83": 727, "%84": 728, "%85": 729, "%86": 730, "[RuH+3]": 731, "%87": 732, "%88": 733, "%89": 734, "%90": 735, "%91": 736, "%92": 737, "[Ca+]": 738, "[Ba+]": 739, "[MnH]": 740, "[ZnH+]": 741, "%93": 742, "%94": 743, "%95": 744, "%96": 745, "%97": 746, "%98": 747, "%99": 748, "0": 749, "[CuH+]": 750, "[TiH+3]": 751, "[VH]": 752, "[ClH2+2]": 753, "[Ge+4]": 754, "[Sb+5]": 755, "[SeH5]": 756, "[AlH+]": 757, "[ClH3+2]": 758, "[TiH]": 759, "[TiH+]": 760, "[ClH3+3]": 761, "[ClH+2]": 762, "[Cd-]": 763, "[PdH]": 764, "[siH+]": 765, "[SiH+]": 766, "[XeH]": 767, "[siH-]": 768, "[KrH]": 769, "[AsH+]": 770, "[AsH2-]": 771, "[SiH4-]": 772, "[SnH2-]": 773, "[YbH2]": 774, "[CrH+2]": 775, "[CuH2-]": 776, "[CoH+]": 777, "*": 778, "[Cf+3]": 779, "[CoH+2]": 780, "[MnH+]": 781, "[CeH]": 782, "[NiH2+2]": 783, "[PoH]": 784, "[SnH-]": 785, "[SiH+2]": 786, "[ClH2+3]": 787, "[SbH3+]": 788, "[AuH3]": 789, "[XeH2]": 790, "[B+]": 791, "[TaH5]": 792, "[FeH4-]": 793, "[FeH6-3]": 794, "[ScH3]": 795, "[Ru-4]": 796, "[SbH2+]": 797, "[RuH-]": 798}