Commit 3971bd3c authored by Bharath Ramsundar's avatar Bharath Ramsundar
Browse files

Batch of changes to fix vanilla multitask networks.

parent 2c0ea7f2

deep_chem/models/deep.py

+9 −3
Original line number Diff line number Diff line
@@ -43,6 +43,7 @@ def process_multitask(paths, task_transforms, splittype="random", 
    Seed used for random splits.

  """

  dataset = load_and_transform_dataset(paths, task_transforms,

			prediction_endpoint,

      weight_positives=weight_positives)

  sorted_targets = sorted(dataset.keys())

  if splittype == "random":
@@ -63,7 +64,9 @@ def process_multitask(paths, task_transforms, splittype="random", 
  #  ensure_balanced(y_test, W_test)

  return (train, X_train, y_train, W_train, test, X_test, y_test, W_test)



def process_singletask(paths, task_transforms, splittype="random", seed=None,

def process_singletask(paths, task_transforms,

		prediction_endpoint,

		splittype="random", seed=None,

    weight_positives=True):

  """Extracts singletask datasets and splits into train/test.
@@ -82,6 +85,7 @@ def process_singletask(paths, task_transforms, splittype="random", seed=None, 
    Seed used for random splits.

  """

  dataset = load_and_transform_dataset(paths, task_transforms,

			prediction_endpoint,

      weight_positives=weight_positives)

  singletask = multitask_to_singletask(dataset)

  arrays = {}
@@ -101,7 +105,7 @@ def process_singletask(paths, task_transforms, splittype="random", seed=None, 
  return arrays





def fit_multitask_mlp(paths, task_types, task_transforms,

def fit_multitask_mlp(paths, task_types, task_transforms, prediction_endpoint,

                      splittype="random", weight_positives=False, **training_params):

  """

  Perform stochastic gradient descent optimization for a keras multitask MLP.
@@ -137,6 +141,7 @@ def fit_multitask_mlp(paths, task_types, task_transforms, 
    print "Mean R^2: %f" % np.mean(np.array(r2s.values()))



def fit_singletask_mlp(paths, task_types, task_transforms,

											 prediction_endpoint,

                       splittype="random", weight_positives=True,

                       num_to_train=None, **training_params):

  """
@@ -154,6 +159,7 @@ def fit_singletask_mlp(paths, task_types, task_transforms, 
    Aggregates keyword parameters to pass to train_multitask_model

  """

  singletasks = process_singletask(paths, task_transforms,

		prediction_endpoint,

    splittype=splittype, weight_positives=weight_positives)

  ret_vals = {}

  aucs, r2s, rms = {}, {}, {}
@@ -163,7 +169,7 @@ def fit_singletask_mlp(paths, task_types, task_transforms, 
  for index, target in enumerate(sorted_targets):

    print "Training model %d" % index

    print "Target %s" % target

    (train, X_train, y_train, W_train, test, X_test, y_test, W_test) = (

    (train, X_train, y_train, W_train), (test, X_test, y_test, W_test) = (

        singletasks[target])

    model = train_multitask_model(X_train, y_train, W_train,

        {target: task_types[target]}, **training_params)

deep_chem/models/deep3d.py

+6 −4
Original line number Diff line number Diff line
@@ -17,7 +17,7 @@ from deep_chem.utils.evaluate import compute_r2_scores 
# code in deep.py

# TODO(rbharath): paths is to handle sharded input pickle files. Might be

# better to use hdf5 datasets like in MSMBuilder

def process_3D_convolutions(paths, task_transforms, seed=None, splittype="random"):

def process_3D_convolutions(paths, task_transforms, prediction_endpoint, seed=None, splittype="random"):

  """Loads 3D Convolution datasets.



  Parameters
@@ -25,7 +25,8 @@ def process_3D_convolutions(paths, task_transforms, seed=None, splittype="random 
  paths: list

    List of paths to convolution datasets.

  """

  dataset = load_and_transform_dataset(paths, task_transforms, datatype="pdbbind")

  dataset = load_and_transform_dataset(paths, task_transforms,

    prediction_endpoint, datatype="pdbbind")

  # TODO(rbharath): Factor this code splitting out into a util function.

  if splittype == "random":

    train, test = train_test_random_split(dataset, seed=seed)
@@ -35,12 +36,13 @@ def process_3D_convolutions(paths, task_transforms, seed=None, splittype="random 
  X_test, y_test, W_test = tensor_dataset_to_numpy(test)

  return (X_train, y_train, W_train, train), (X_test, y_test, W_test, test)



def fit_3D_convolution(paths, task_types, task_transforms, axis_length=32, **training_params):

def fit_3D_convolution(paths, task_types, task_transforms, prediction_endpoint,

    axis_length=32, **training_params):

  """

  Perform stochastic gradient descent for a 3D CNN.

  """

  (X_train, y_train, W_train, train), (X_test, y_test, W_test, test) = process_3D_convolutions(

    paths, task_transforms)

    paths, task_transforms, prediction_endpoint)



  print "np.shape(X_train): " + str(np.shape(X_train))

  print "np.shape(y_train): " + str(np.shape(y_train))

deep_chem/scripts/launch_models.py

+9 −3
Original line number Diff line number Diff line
@@ -25,6 +25,8 @@ def parse_args(input_args=None): 
  parser.add_argument("--splittype", type=str, default="scaffold",

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

                       help="Type of cross-validation data-splitting.")

  parser.add_argument("--prediction-endpoint", type=str, default="IC50",

                       help="Name of measured endpoint to predict.")

  parser.add_argument("--n-hidden", type=int, default=500,

                      help="Number of hidden neurons for NN models.")

  parser.add_argument("--learning-rate", type=float, default=0.01,
@@ -53,7 +55,6 @@ def main(): 
  args = parse_args()

  paths = {}





  for dataset, path in zip(args.datasets, args.paths):

    paths[dataset] = path
@@ -61,20 +62,25 @@ def main(): 


  if args.model == "singletask_deep_network":

    fit_singletask_mlp(paths.values(), task_types, task_transforms,

      splittype=args.splittype, n_hidden=args.n_hidden,

      prediction_endpoint=args.prediction_endpoint,

      splittype=args.splittype, 

      n_hidden=args.n_hidden,

      learning_rate=args.learning_rate, dropout=args.dropout,

      nb_epoch=args.n_epochs, decay=args.decay, batch_size=args.batch_size,

      validation_split=args.validation_split,

      weight_positives=args.weight_positives, num_to_train=args.num_to_train)

  elif args.model == "multitask_deep_network":

    fit_multitask_mlp(paths.values(), task_types, task_transforms,

      splittype=args.splittype, n_hidden=args.n_hidden, learning_rate =

      prediction_endpoint=args.prediction_endpoint,

      splittype=args.splittype,

      n_hidden=args.n_hidden, learning_rate =

      args.learning_rate, dropout = args.dropout, batch_size=args.batch_size,

      nb_epoch=args.n_epochs, decay=args.decay,

      validation_split=args.validation_split,

      weight_positives=args.weight_positives)

  elif args.model == "3D_cnn":

    fit_3D_convolution(paths.values(), task_types, task_transforms,

        prediction_endpoint=args.prediction_endpoint,

        axis_length=args.axis_length, nb_epoch=args.n_epochs,

        batch_size=args.batch_size)

  else:

deep_chem/scripts/process_bace.sh

+1 −1
Original line number Diff line number Diff line 
# Usage ./process_bace.sh INPUT_SDF_FILE

python -m deep_chem.scripts.process_dataset --input_file $1 --input-type sdf --fields Name smiles pIC50 --field-types string string concentration --name BACE --out /tmp/
 
python -m deep_chem.scripts.process_dataset --input-file $1 --input-type sdf --fields Name smiles pIC50 --field-types string string concentration --name BACE --out /tmp/

deep_chem/scripts/process_dataset.py

+21 −3
Original line number Diff line number Diff line
@@ -105,8 +105,20 @@ def get_rows(input_file, input_type): 
      df = pickle.load(f)

    return df.iterrows()

  elif input_type == "sdf":

    if ".gz" in input_file:

      print "gzipped"

      with gzip.open(input_file) as f:

      return Chem.ForwardSDMolSupplier(f)

        supp = Chem.ForwardSDMolSupplier(f)

        mols = [mol for mol in supp if mol is not None]

      print "len(mols): " + str(len(mols))

      return mols

    else:

      print "non-gzipped"

      with open(input_file) as f:

        supp  = Chem.ForwardSDMolSupplier(f)

        mols = [mol for mol in supp if mol is not None]

      print "len(mols): " + str(len(mols))

      return mols



def get_row_data(row, input_type, fields, field_types):

  """Extract information from row data."""
@@ -133,6 +145,7 @@ def get_row_data(row, input_type, fields, field_types): 
        row_data[ind] = None

      else:

        row_data[ind] = mol.GetProp(field)

    return row_data



def process_field(data, field_type):

  """Parse data in a field."""
@@ -141,7 +154,11 @@ def process_field(data, field_type): 
  elif field_type == "float":

    return parse_float_input(data)

  elif field_type == "concentration":

    return parse_float_input(data) / 1e-6

    fl = parse_float_input(data)

    if fl is not None:

      return parse_float_input(data) / 1e-7

    else:

      return None

  elif field_type == "list-string":

    return data.split(",")

  elif field_type == "list-float":
@@ -155,6 +172,7 @@ def generate_targets(input_file, input_type, fields, field_types, out_pkl, 
  rows, mols, smiles = [], [], SmilesGenerator()

  for row_index, raw_row in enumerate(get_rows(input_file, input_type)):

    print row_index

    print raw_row

    # Skip row labels.

    if row_index == 0 or raw_row is None:

      continue

CRA Git | Maintained and supported by SUSTech CRA and CCSE