Made a few changes to make it easier to invoke model from shell script wrapper.

  train_test_cmd.add_argument("--input-transforms", nargs="+", default=[],

      choices=["normalize-and-truncate"],

      help="Transforms to apply to input data.")

  train_test_cmd.add_argument("--output-transforms", nargs="+", default=[],

      choices=["log", "normalize"],

      help="Transforms to apply to output data.")

  train_test_cmd.add_argument("--feature-types", nargs="+", required=1,

      help="Types of featurizations to use. Each featurization must correspond\n"

           "to subdirectory in generated data directory.")

  train_test_cmd.add_argument("--output-transforms", type=str, default="",

      help="Comma-separated list (no spaces) of transforms to apply to output data.\n"

           "Supported transforms are 'log' and 'normalize'. 'None' will be taken\n"

           "to mean no transforms are required.")

  train_test_cmd.add_argument("--feature-types", type=str, required=1,

      help="Comma-separated list (no spaces) of types of featurizations to use.\n"

           "Each featurization must correspond to subdirectory in generated\n"

           "data directory.")

  train_test_cmd.add_argument("--paths", nargs="+", required=1,

      help="Paths to input datasets.")

  train_test_cmd.add_argument("--splittype", type=str, default="scaffold",

           "for user convenience.")

  model_cmd.add_argument("--skip-featurization", action="store_true",

      help="If set, skip the featurization step.")

  model_cmd.add_argument("--feature-types", nargs="+", required=1,

      help="Types of featurizations to use. Each featurization must correspond\n"

           "to subdirectory in generated data directory.")

  add_featurize_group(model_cmd)

  train_test_group = model_cmd.add_argument_group("train_test_group")

  train_test_group.add_argument("--input-transforms", nargs="+", default=[],

      choices=["normalize-and-truncate"],

      help="Transforms to apply to input data.")

  train_test_group.add_argument("--output-transforms", nargs="+", default=[],

      choices=["log", "normalize"],

      help="Transforms to apply to output data.")

  train_test_group.add_argument("--output-transforms", type=str, default="",

      help="Comma-separated list (no spaces) of transforms to apply to output data.\n"

           "Supported transforms are log and normalize.")

  train_test_group.add_argument("--mode", default="singletask",

      choices=["singletask", "multitask"],

      help="Type of model being built.")

  train_test_group.add_argument("--feature-types", type=str, required=1,

      help="Comma-separated list (no spaces) of types of featurizations to use.\n"

           "Each featurization must correspond to subdirectory in generated\n"

           "data directory.")

  train_test_group.add_argument("--splittype", type=str, default="scaffold",

      choices=["scaffold", "random", "specified"],

      help="Type of train/test data-splitting. 'scaffold' uses Bemis-Murcko scaffolds.\n"

  print "Perform train-test split"

  paths = [data_dir]

  weight_positives = False  # Hard coding this for now

  train_out = os.path.join(args.out, "%s-train.pkl.gz" % args.name)

  test_out = os.path.join(args.out, "%s-test.pkl.gz" % args.name)

  train_out = os.path.join(data_dir, "%s-train.pkl.gz" % args.name)

  test_out = os.path.join(data_dir, "%s-test.pkl.gz" % args.name)

  _train_test_input(paths, args.output_transforms,

      args.input_transforms, args.feature_types, args.splittype,

      weight_positives, args.mode, train_out, test_out)

  print "+++++++++++++++++++++++++++++++++"

  print "Eval Model on Train"

  print "-------------------"

  csv_out_train = os.path.join(data_dir, "%s-train.csv" % args.model)

  csv_out_test = os.path.join(data_dir, "%s-test.csv" % args.model)

  csv_out_train = os.path.join(data_dir, "%s-train.csv" % args.name)

  csv_out_test = os.path.join(data_dir, "%s-test.csv" % args.name)

  compute_aucs, compute_recall, compute_accuracy, compute_matthews_corrcoef = (

    False, False, False, False)

  compute_r2s, compute_rms = False, False 

    feature_types, splittype, weight_positives, mode, train_out, test_out):

  """Saves transformed model."""

  targets = get_target_names(paths)

  if output_transforms == "" or output_transforms == "None":

    output_transforms = []

  else:

    output_transforms = output_transforms.split(",")

  output_transforms_dict = {target: output_transforms for target in targets}

  feature_types = feature_types.split(",")

  train_dict, test_dict= process_datasets(paths,

      input_transforms, output_transforms_dict, feature_types=feature_types, 

      splittype=splittype, weight_positives=weight_positives,

  save_model(models, modeltype, saved_out)

def get_model_type(model):

  if model in ["singletask_deep_network", "multitask_deep_network", "3D_cnn"]:

    modeltype = "keras"

  if model in ["singletask_deep_network", "multitask_deep_network"]:

    modeltype = "keras-graph"

  elif model in ["3D_cnn"]:

    modeltype = "keras-sequential"

  elif model == "neural_fingerprint":

    modeltype = "autograd"

  else:

    modeltype = "sklearn"

  return modeltype

    return "joblib"

  elif modeltype == "autograd":

    return "pkl.gz"

  elif modeltype == "keras":

  elif modeltype == "keras-graph" or modeltype == "keras-sequential":

    return "h5"

def eval_trained_model(args):