CODEWITHSUNDEEP
pythontensorflowanaconda
Rafa
Rafa

Reputation: 3043

Anaconda & Tensorflow & Skflow: dnn() got an unexpected keyword argument 'keep_prob'

When running this example in Anaconda 2.7:

import tensorflow.contrib.learn as skflow
def DNN_model(X, y):
    """This is DNN with 50, 20, 10 hidden layers, and dropout of 0.5 probability."""
    layers = skflow.ops.dnn(X, [50, 30, 10], keep_prob=0.5)
    return skflow.models.logistic_regression(layers, y)
clf = skflow.TensorFlowEstimator(model_fn=DNN_model, n_classes=3)

It dumps the following problem:

dnn() got an unexpected keyword argument 'keep_prob'

Tensorflow in Anaconda was installed using:

conda install -c jjhelmus tensorflow=0.9.0

Any idea what failed?

Upvotes: 2

Views: 1346

Answers (2)

yu yang Jian
yu yang Jian

Reputation: 7171

In my case, keep_prob change to rate,

# tf.nn.dropout(D_Hidden, keep_prob=0.8)
tf.nn.dropout(D_Hidden, rate=0.2)

Please use rate instead of keep_prob . Rate should be set to rate = 1 - keep_prob. https://www.tensorflow.org/api_docs/python/tf/compat/v1/nn/dropout

Upvotes: 0

Rafa
Rafa

Reputation: 3043

According to the repository it seems this version is depreciated:

learn.ops.dnn is deprecated,please use contrib.layers.dnn.

However, different arguments should be passed:

def dnn(tensor_in, hidden_units, activation=nn.relu, dropout=None):
  """Creates fully connected deep neural network subgraph.
  This is deprecated. Please use contrib.layers.dnn instead.
  Args:
    tensor_in: tensor or placeholder for input features.
    hidden_units: list of counts of hidden units in each layer.
    activation: activation function between layers. Can be None.
    dropout: if not None, will add a dropout layer with given probability.
  Returns:
    A tensor which would be a deep neural network.
  """

Upvotes: 3

Related Questions