Commit a7fca56e authored by Bharath Ramsundar's avatar Bharath Ramsundar
Browse files

Dropout fix and debugging

parent 9ab2ac00

deepchem/metrics/__init__.py

+8 −6
Original line number Diff line number Diff line
@@ -104,7 +104,7 @@ class Metric(object): 
  """Wrapper class for computing user-defined metrics."""



  def __init__(self, metric, task_averager=None, name=None, threshold=None,

               verbosity=None, mode=None, compute_energy_metric=False):

               verbosity="high", mode=None, compute_energy_metric=False):

    """

    Args:

      metric: function that takes args y_true, y_pred (in that order) and
@@ -127,10 +127,12 @@ class Metric(object): 
    self.verbosity = verbosity

    self.threshold = threshold

    if mode is None:

      if self.metric.__name__ in ["roc_auc_score", "matthews_corrcoef", "recall_score",

                       "accuracy_score", "kappa_score", "precision_score"]:

      if self.metric.__name__ in ["roc_auc_score", "matthews_corrcoef",

                                  "recall_score", "accuracy_score",

                                  "kappa_score", "precision_score"]:

        mode = "classification"

      elif self.metric.__name__ in ["pearson_r2_score", "r2_score", "mean_squared_error",

      elif self.metric.__name__ in ["pearson_r2_score", "r2_score",

                                    "mean_squared_error",

                                    "mean_absolute_error", "rms_score",

                                    "mae_score"]:

        mode = "regression"

deepchem/models/models.py

+11 −1
Original line number Diff line number Diff line
@@ -162,8 +162,18 @@ class Model(object): 
    """

    y_preds = []

    n_tasks = self.get_num_tasks()

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

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

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

    ind = 0

    for (X_batch, _, _, ids_batch) in dataset.iterbatches(

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

        batch_size, deterministic=True):

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

      if ind == 0:

        print("ids_batch[0]")

        print(ids_batch[0])

      ind += 1

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

      n_samples = len(X_batch)

      y_pred_batch = self.predict_on_batch(X_batch, pad_batch=pad_batches)

      # Discard any padded predictions

deepchem/models/tensorflow_models/__init__.py

+5 −26
Original line number Diff line number Diff line
@@ -243,7 +243,7 @@ class TensorflowGraphModel(Model): 


      return loss 



  def fit(self, dataset, nb_epoch=10, pad_batches=False, shuffle=False,

  def fit(self, dataset, nb_epoch=10, pad_batches=False, 

          max_checkpoints_to_keep=5, log_every_N_batches=50, **kwargs):

    """Fit the model.
@@ -258,9 +258,6 @@ class TensorflowGraphModel(Model): 
    ############################################################## TIMING

    time1 = time.time()

    ############################################################## TIMING

    n_datapoints = len(dataset)

    batch_size = self.batch_size

    step_per_epoch = np.ceil(float(n_datapoints)/batch_size)

    log("Training for %d epochs" % nb_epoch, self.verbosity)

    with self.train_graph.graph.as_default():

      train_op = self.get_training_op(
@@ -272,14 +269,11 @@ class TensorflowGraphModel(Model): 
        saver.save(sess, self._save_path, global_step=0)

        for epoch in range(nb_epoch):

          avg_loss, n_batches = 0., 0

          if shuffle:

            log("About to shuffle dataset before epoch start.", self.verbosity)

            dataset.shuffle()

          for ind, (X_b, y_b, w_b, ids_b) in enumerate(

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

              ## hardcode pad_batches=True to work around limitations in Tensorflow

              #dataset.iterbatches(batch_size, pad_batches=True)):

              dataset.iterbatches(batch_size, pad_batches=False)):

              dataset.iterbatches(self.batch_size, pad_batches=pad_batches)):

              #dataset.iterbatches(batch_size, pad_batches=pad_batches)):

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

            if ind % log_every_N_batches == 0:
@@ -288,9 +282,7 @@ class TensorflowGraphModel(Model): 
            feed_dict = self.construct_feed_dict(X_b, y_b, w_b, ids_b)

            fetches = self.train_graph.output + [

                train_op, self.train_graph.loss]

            fetched_values = sess.run(

                fetches,

                feed_dict=feed_dict)

            fetched_values = sess.run(fetches, feed_dict=feed_dict)

            output = fetched_values[:len(self.train_graph.output)]

            loss = fetched_values[-1]

            avg_loss += loss
@@ -419,7 +411,6 @@ class TensorflowGraphModel(Model): 
      return

    with self.eval_graph.graph.as_default():

      last_checkpoint = self._find_last_checkpoint()



      # TODO(rbharath): Is setting train=False right here?

      saver = tf.train.Saver()

      saver.restore(self._get_shared_session(train=False),
@@ -448,13 +439,7 @@ class TensorflowClassifier(TensorflowGraphModel): 
  Subclasses must set the following attributes:

    output: logits op(s) used for computing classification loss and predicted

      class probabilities for each task.



  Class attributes:

    default_metrics: List of metrics to compute by default.

  """



  default_metrics = ['auc']



  def get_task_type(self):

    return "classification"
@@ -603,13 +588,7 @@ class TensorflowRegressor(TensorflowGraphModel): 
  Subclasses must set the following attributes:

    output: Op(s) used for computing regression loss and predicted regression

      outputs for each task.



  Class attributes:

    default_metrics: List of metrics to compute by default.

  """



  default_metrics = ['r2']



  def get_task_type(self):

    return "regressor"
@@ -677,7 +656,7 @@ class TensorflowRegressor(TensorflowGraphModel): 
    

    if not self._restored_model:

      self.restore()

    with self.train_graph.graph.as_default():

    with self.eval_graph.graph.as_default():



      # run eval data through the model

      n_tasks = self.n_tasks

deepchem/models/tensorflow_models/fcnet.py

+1 −1
Original line number Diff line number Diff line
@@ -142,7 +142,7 @@ class TensorflowMultiTaskRegressor(TensorflowRegressor): 
                stddev=weight_init_stddevs[i]),

            bias_init=tf.constant(value=bias_init_consts[i],

                                  shape=[layer_sizes[i]])))

        layer = model_ops.dropout(layer, dropouts[i])

        layer = model_ops.dropout(layer, dropouts[i], training)

        prev_layer = layer

        prev_layer_size = layer_sizes[i]

deepchem/utils/evaluate.py

+40 −6
Original line number Diff line number Diff line
@@ -35,8 +35,11 @@ class Evaluator(object): 
  def __init__(self, model, dataset, transformers, verbosity=False):

    self.model = model

    self.dataset = dataset

    self.output_transformers = [

        transformer for transformer in transformers if transformer.transform_y]

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

    #self.output_transformers = [

    #    transformer for transformer in transformers if transformer.transform_y]

    self.transformers = transformers

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

    self.task_names = dataset.get_task_names()

    self.verbosity = verbosity
@@ -71,19 +74,50 @@ class Evaluator(object): 
    Computes statistics of model on test data and saves results to csv.

    """

    y = self.dataset.y

    y = undo_transforms(y, self.output_transformers)

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

    #print("self.output_transformers")

    #print(self.output_transformers)

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

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

    #y = undo_transforms(y, self.output_transformers)

    y = undo_transforms(y, self.transformers)

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

    w = self.dataset.w



    if not len(metrics):

      return {}

    else:

      mode = metrics[0].mode

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

    print("mode")

    print(mode)

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

    if mode == "classification":

      y_pred = self.model.predict_proba(self.dataset, self.output_transformers)

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

      #y_pred = self.model.predict_proba(self.dataset, self.output_transformers)

      #y_pred_print = self.model.predict(

      #    self.dataset, self.output_transformers).astype(int)

      y_pred = self.model.predict_proba(self.dataset, self.transformers)

      y_pred_print = self.model.predict(

          self.dataset, self.output_transformers).astype(int)

          self.dataset, self.transformers).astype(int)

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

    else:

      y_pred = self.model.predict(self.dataset, self.output_transformers)

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

      #y_pred = self.model.predict(self.dataset, self.output_transformers)

      y_pred = self.model.predict(self.dataset, self.transformers)

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

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

      print("y_pred.shape")

      print(y_pred.shape)

      print("y_pred[:1]")

      print(y_pred[:1])

      print("y[:1]")

      print(y[:1])

      raw_y = self.model.predict_on_batch(self.dataset.X[:1])

      raw_y = undo_transforms(raw_y, self.transformers)

      print("raw_y")

      print(raw_y)

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

      y_pred_print = y_pred

    multitask_scores = {}

CRA Git | Maintained and supported by SUSTech CRA and CCSE