Commit d8575698 authored by Bharath Ramsundar's avatar Bharath Ramsundar
Browse files

Some cleanup and minor-bugfixes on tox21 examples.

parent 8216fd80

.travis.yml

+0 −2
Original line number Diff line number Diff line
@@ -29,8 +29,6 @@ install: 
- conda install -c omnia mdtraj 

- python setup.py install

- conda install -c https://conda.anaconda.org/jjhelmus tensorflow

#- pip install --upgrade --no-deps theano

#- pip install keras

script:

- nosetests -v deepchem --nologcapture

after_success:

deepchem/datasets/bace_datasets.py

+4 −11
Original line number Diff line number Diff line
@@ -29,8 +29,6 @@ def load_bace(mode="regression", transform=True, split="20-80"): 
  elif split == "80-20":

    dataset_file = os.path.join(

        current_dir, "../../datasets/rev8020split_desc.csv")

  print("dataset_file")

  print(dataset_file)

  dataset = load_from_disk(dataset_file)

  num_display = 10

  pretty_columns = (
@@ -85,11 +83,6 @@ def load_bace(mode="regression", transform=True, split="20-80"): 
  train_samples, valid_samples, test_samples = splitter.train_valid_test_split(

      featurized_samples, train_dir, valid_dir, test_dir,

      reload=reload)

  ######################### DEBUG

  #print("bace_datasets")

  #print("len(train_samples), len(valid_samples), len(test_samples)")

  #print(len(train_samples), len(valid_samples), len(test_samples))

  ######################### DEBUG



  #NOTE THE RENAMING:

  if split == "20-80":
@@ -137,8 +130,8 @@ def load_bace(mode="regression", transform=True, split="20-80"): 
      transformer.transform(valid_dataset)

  for transformer in transformers:

      transformer.transform(test_dataset)

  #for transformer in transformers:

  #    transformer.transform(crystal_dataset)

  for transformer in transformers:

      transformer.transform(crystal_dataset)



  #return (bace_tasks, train_dataset, valid_dataset, test_dataset, crystal_dataset, transformers)

  return (bace_tasks, train_dataset, valid_dataset, test_dataset, crystal_dataset, output_transformers)
 
  return (bace_tasks, train_dataset, valid_dataset, test_dataset,

          crystal_dataset, output_transformers)

deepchem/models/__init__.py

+0 −12
Original line number Diff line number Diff line
@@ -143,18 +143,6 @@ class Model(object): 
      y_preds.append(y_pred_batch)

    y_pred = np.vstack(y_preds)

  

    ################ DEBUG

    #y_plain = []

    #for (X_batch, y_batch, w_batch, ids_batch) in dataset.iterbatches(batch_size):

    #  y_pred_batch = np.reshape(self.predict_on_batch(X_batch), y_batch.shape)

    #  y_plain.append(y_pred_batch)

    #y_plain_pred = np.vstack(y_plain)

    #y = dataset.get_labels()

    #import sklearn

    #print("sklearn.metrics.r2_score(y, y_plain_pred)")

    #print(sklearn.metrics.r2_score(y, y_plain_pred))

    ################ DEBUG

  

    # The iterbatches does padding with zero-weight examples on the last batch.

    # Remove padded examples.

    n_samples, n_tasks = len(dataset), len(self.tasks)

examples/tox21/tox21_sklearn_models.py

+2 −29
Original line number Diff line number Diff line
@@ -69,7 +69,7 @@ valid_scores = {} 


print("Using following tasks")

print(tox21_tasks)

##### CHANGE IF N_TASKS IS CHANGED!



## This is for good debug (to make sure nasty state isn't being passed around)

if os.path.exists(feature_dir):

  shutil.rmtree(feature_dir)
@@ -117,39 +117,12 @@ valid_dataset = Dataset.from_numpy(valid_dir, X_valid, y_valid, 
transformers = []



# Fit models

tox21_task_types = {task: "Classification" for task in tox21_tasks}

tox21_task_types = {task: "classification" for task in tox21_tasks}



classification_metric = Metric(metrics.roc_auc_score, np.mean,

                               verbosity=verbosity,

                               mode="classification")



####################### DEBUG

y_train = np.reshape(y_train, (len(y_train), n_tasks))

print("w_valid")

print(w_valid)

for task, _ in enumerate(tox21_tasks):

  print("BASELINE FOR TASK %s" % str(task))

  model = RandomForestClassifier(class_weight="balanced",

                                 n_estimators=500)

  y_train_task = y_train[:, task]

  y_valid_task = y_valid[:, task]

  w_valid_task = w_valid[:, task]

  model.fit(X_train, y_train_task)

  #y_valid_task = metrics.to_one_hot(y_valid_task)

  y_valid_task = y_valid_task[:, np.newaxis]

  y_pred_task = model.predict_proba(X_valid)

  y_pred_task = y_pred_task[:, np.newaxis]

  w_valid_task = w_valid_task[:, np.newaxis]

  print("y_valid_task.shape, y_pred_task.shape, w_valid_task.shape")

  print(y_valid_task.shape, y_pred_task.shape, w_valid_task.shape)

  print("classification_metric.compute_metric(y_valid_task, y_pred_task, w_valid_task)")

  print(classification_metric.compute_metric(y_valid_task, y_pred_task, w_valid_task))

  #print("metrics.roc_auc_score(y_valid_task, y_pred_task)")

  #print(metrics.roc_auc_score(y_valid_task, y_pred_task))

  print("----------------------")

####################### DEBUG





params_dict = { 

    "batch_size": None,

    "data_shape": train_dataset.get_data_shape(),

examples/tox21/tox21_tf_models.py

+0 −13
Original line number Diff line number Diff line
@@ -65,13 +65,8 @@ print("About to featurize Tox21 dataset.") 
featurizers = [CircularFingerprint(size=1024)]

tox21_tasks = ['NR-AR', 'NR-AR-LBD', 'NR-AhR', 'NR-Aromatase', 'NR-ER', 'NR-ER-LBD',

               'NR-PPAR-gamma', 'SR-ARE', 'SR-ATAD5', 'SR-HSE', 'SR-MMP', 'SR-p53']

# For debugging purposes

#tox21_tasks = tox21_tasks[0:1]

#tox21_tasks = tox21_tasks[0:2]

all_valid_scores = {}



print("Using following tasks")

print(tox21_tasks)

# This is for good debug (to make sure nasty state isn't being passed around)

if os.path.exists(feature_dir):

  shutil.rmtree(feature_dir)
@@ -86,8 +81,6 @@ featurized_samples = featurizer.featurize( 


# Generate datasets

print("About to create datasets")

print("tox21_tasks")

print(tox21_tasks)



# This is for good debug (to make sure nasty state isn't being passed around)

if os.path.exists(full_dir):
@@ -107,11 +100,6 @@ ids_train, ids_valid = ids[:num_train], ids[num_train:] 
# Not sure if we need to constantly delete these directories...

if os.path.exists(train_dir):

  shutil.rmtree(train_dir)

print("tox21_tf_models.py")

print("type(y_train)")

print(type(y_train))

print("type(X_train)")

print(type(X_train))

train_dataset = Dataset.from_numpy(train_dir, X_train, y_train,

                                   w_train, ids_train, tox21_tasks)
@@ -131,7 +119,6 @@ classification_metric = Metric(metrics.roc_auc_score, np.mean, 
                               mode="classification")

params_dict = { 

    "batch_size": 32,

    #"batch_size": 5120,

    "nb_epoch": 50,

    "data_shape": train_dataset.get_data_shape(),

    "layer_sizes": [1000],

CRA Git | Maintained and supported by SUSTech CRA and CCSE