solved some bugs

                          axis_length, axis_length, axis_length)

      learning_rate = model_params["learning_rate"]

      print("learning rate = %f" % learning_rate)

      loss_function = model_params["loss_function"]

         # number of convolutional filters to use at each layer

  split_df = metadata_df.loc[metadata_df['split'] == split]

  nb_batch = split_df.shape[0]

  MAX_GPU_RAM = float(691007488/50)

  MAX_GPU_RAM = float(691007488/100)

  for i, row in split_df.iterrows():

    print("Evaluating on %s batch %d out of %d" % (split, i+1, nb_batch))

  performance_df = pd.DataFrame(columns=colnames)

  y_means = pred_y_df.iterrows().next()[1]["y_means"]

  print("y_means")

  print(y_means)

  y_stds = pred_y_df.iterrows().next()[1]["y_stds"]

  print("y_stds")

  print(y_stds)

  for i, task_name in enumerate(task_names):

    y = pred_y_df[task_name]

  _, row = metadata_df.iterrows().next()

  return row['task_names']

def fit_model(model_name, model_params, model_dir, data_dir):

def fit_model(model_name, model_params, model_dir, data_dir, nb_epoch=25):

  """Builds model from featurized data."""

  task_type = get_task_type(model_name)

  metadata_filename = get_metadata_filename(data_dir)

  print(model)

  train_metadata = metadata_df.loc[metadata_df['split'] =="train"]

  for epoch in range(0, nb_epoch):

    nb_batch = train_metadata.shape[0]

  MAX_GPU_RAM = float(691007488/50)

    MAX_GPU_RAM = float(691007488/100)

    for i, row in train_metadata.iterrows():

      print("Training on batch %d out of %d" % (i+1, nb_batch))

      X = load_sharded_dataset(row['X-transformed'])

          model.fit_on_batch(X_batch, y_batch, w_batch)

      else:

        model.fit_on_batch(X, y, w)

    print("Completed fitting epoch %d" % epoch)

  save_model(model, model_name, model_dir)

def undo_normalization(y, y_means, y_stds):

  """Undo the applied normalization transform."""

  y = y * y_means + y_stds

  return y * y_means + y_stds

  y = y * y_stds + y_means

  return y

def undo_transform(y, y_means, y_stds, output_transforms):

  """Undo transforms on y_pred, W_pred."""

  elif output_transforms == ["log"]:

    return np.exp(y)

  elif output_transforms == ["normalize"]:

    print("Undoing normalization.")

    return undo_normalization(y, y_means, y_stds)

  elif output_transforms == ["log", "normalize"]:

    return np.exp(undo_normalization(y, y_means, y_stds))