Getting "Tensorflow %s is not valid scope name error" while I am trying to create a model for kaggle competition

Question

Here are my complete Code and Traceback. This is the starter code for my ml model. There are going to be a lot of additions to it.

import pandas as pd
import tensorflow as tf
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn import metrics
from IPython import display
from tensorflow.python.data import Dataset
tf.logging.set_verbosity(tf.logging.ERROR)
pd.options.display.max_rows = 5
pd.options.display.float_format = '{:.1f}'.format

housing_data = pd.read_csv("train.csv")
housing_data = housing_data.reindex(
np.random.permutation(housing_data.index))
housing_data = pd.get_dummies(
housing_data).dropna()

Preprocessing the features. More work is to be done here.

def preprocess_features(housing_data):
    selected_features = housing_data
    selected_features = selected_features.drop(columns = "SalePrice")
    processed_features = selected_features.copy()
    return processed_features

def preprocess_target(housing_data):
    output_target = pd.DataFrame()
    output_target["SalePrice"] = (housing_data.SalePrice / 1000.0)
    return output_target

training_examples = preprocess_features(housing_data.head(900))
training_targets = preprocess_target(housing_data.head(900))

validation_examples = preprocess_features(housing_data.tail(221))
validation_targets = preprocess_target(housing_data.tail(221))

def construct_feature_columns(input_features):
    '''
    Returns the set of feature columns for tf.estimator classifiers and regressors
    '''
    return set([tf.feature_column.numeric_column(my_feature) for my_feature in input_features])


def my_input_fn(features, targets, batch_size = 1, shuffle = True, num_epochs = None):
    #convert the pandas dataframe into a numpy array

    features = {key:np.array(value) for key,value in dict(features).items()}

    #create the dataset
    ds = Dataset.from_tensor_slices((features,targets))
    ds = ds.batch(batch_size).repeat(num_epochs)

    #shuffle the data
    if shuffle:
        ds = ds.shuffle(1000)

    #return the features and targets tuple for next iteration
    features,labels= 
ds.make_one_shot_iterator().get_next()
    return features,labels

The Linear Classifier

def train_linear_classifier_model(
    learning_rate,
    regularization_strength,
    steps,
    batch_size,
    training_examples,
    training_targets,
    validation_examples,
    validation_targets
):
    periods = 10
    steps_per_period = steps / periods

    my_optimizer = tf.train.FtrlOptimizer(learning_rate=learning_rate, l1_regularization_strength=regularization_strength)
    my_optimizer = tf.contrib.estimator.clip_gradients_by_norm(my_optimizer, 5.0)
    linear_classifier = tf.estimator.LinearClassifier(
      feature_columns=construct_feature_columns(training_examples),
      optimizer=my_optimizer
  )


    training_input_fn = lambda: my_input_fn(training_examples, 
                                          training_targets["SalePrice"], 
                                          batch_size=batch_size)
    predict_training_input_fn = lambda: my_input_fn(training_examples, 
                                                  training_targets["SalePrice"], 
                                                  num_epochs=1, 
                                                  shuffle=False)
    predict_validation_input_fn = lambda: my_input_fn(validation_examples, 
                                                    validation_targets["SalePrice"], 
                                                    num_epochs=1, 
                                                    shuffle=False)


    print("Training model...")
    print("LogLoss (on validation data):")
    training_log_losses = []
    validation_log_losses = []
    for period in range (0, periods):
        linear_classifier.train(
        input_fn=training_input_fn,
        steps=steps_per_period
    )
    # Take a break and compute predictions.
    training_probabilities = linear_classifier.predict(input_fn=predict_training_input_fn)
    training_probabilities = np.array([item['probabilities'] for item in training_probabilities])

    validation_probabilities = linear_classifier.predict(input_fn=predict_validation_input_fn)
    validation_probabilities = np.array([item['probabilities'] for item in validation_probabilities])

    # Compute training and validation loss.
    training_log_loss = metrics.log_loss(training_targets, training_probabilities)
    validation_log_loss = metrics.log_loss(validation_targets, validation_probabilities)
    # Occasionally print the current loss.
    print("  period %02d : %0.2f" % (period, validation_log_loss))
    # Add the loss metrics from this period to our list.
    training_log_losses.append(training_log_loss)
    validation_log_losses.append(validation_log_loss)
    print("Model training finished.")

  # Output a graph of loss metrics over periods.
    plt.ylabel("LogLoss")
    plt.xlabel("Periods")
    plt.title("LogLoss vs. Periods")
    plt.tight_layout()
    plt.plot(training_log_losses, label="training")
    plt.plot(validation_log_losses, label="validation")
    plt.legend()

    return linear_classifier

linear_classifier = train_linear_classifier_model(
    learning_rate=0.1,
    regularization_strength=0.1,
    steps=300,
    batch_size=100,
    training_examples=training_examples,
    training_targets=training_targets,
    validation_examples=validation_examples,
    validation_targets = validation_targets)

And here is my Traceback

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-75-f9d203769761> in <module>()
      7     training_targets=training_targets,
      8     validation_examples=validation_examples,
----> 9     validation_targets = validation_targets)

<ipython-input-74-e1dbd56d9615> in train_linear_classifier_model(learning_rate, regularization_strength, steps, batch_size, training_examples, training_targets, validation_examples, validation_targets)
     40         linear_classifier.train(
     41         input_fn=training_input_fn,
---> 42         steps=steps_per_period
     43     )
     44     # Take a break and compute predictions.

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\estimator.py in train(self, input_fn, hooks, steps, max_steps, saving_listeners)
    350 
    351     saving_listeners = _check_listeners_type(saving_listeners)
--> 352     loss = self._train_model(input_fn, hooks, saving_listeners)
    353     logging.info('Loss for final step: %s.', loss)
    354     return self

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\estimator.py in _train_model(self, input_fn, hooks, saving_listeners)
    810       worker_hooks.extend(input_hooks)
    811       estimator_spec = self._call_model_fn(
--> 812           features, labels, model_fn_lib.ModeKeys.TRAIN, self.config)
    813 
    814       if self._warm_start_settings:

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\estimator.py in _call_model_fn(self, features, labels, mode, config)
    791 
    792     logging.info('Calling model_fn.')
--> 793     model_fn_results = self._model_fn(features=features, **kwargs)
    794     logging.info('Done calling model_fn.')
    795 

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\canned\linear.py in _model_fn(features, labels, mode, config)
    314           optimizer=optimizer,
    315           partitioner=partitioner,
--> 316           config=config)
    317 
    318     super(LinearClassifier, self).__init__(

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\canned\linear.py in _linear_model_fn(features, labels, mode, head, feature_columns, optimizer, partitioner, config)
    155     logit_fn = _linear_logit_fn_builder(
    156         units=head.logits_dimension, feature_columns=feature_columns)
--> 157     logits = logit_fn(features=features)
    158 
    159     def _train_op_fn(loss):

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\estimator\canned\linear.py in linear_logit_fn(features)
     96         feature_columns=feature_columns,
     97         units=units,
---> 98         cols_to_vars=cols_to_vars)
     99     bias = cols_to_vars.pop('bias')
    100     if units > 1:

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\feature_column\feature_column.py in linear_model(features, feature_columns, units, sparse_combiner, weight_collections, trainable, cols_to_vars)
    422     for column in sorted(feature_columns, key=lambda x: x.name):
    423       with variable_scope.variable_scope(
--> 424           None, default_name=column._var_scope_name):  # pylint: disable=protected-access
    425         ordered_columns.append(column)
    426         weighted_sum = _create_weighted_sum(

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in __enter__(self)
   1901 
   1902     try:
-> 1903       return self._enter_scope_uncached()
   1904     except:
   1905       if self._graph_context_manager is not None:

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in _enter_scope_uncached(self)
   2001           self._default_name)
   2002       try:
-> 2003         current_name_scope_name = current_name_scope.__enter__()
   2004       except:
   2005         current_name_scope.__exit__(*sys.exc_info())

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\framework\ops.py in __enter__(self)
   5619       try:
   5620         self._name_scope = g.name_scope(self._name)
-> 5621         return self._name_scope.__enter__()
   5622       except:
   5623         self._g_manager.__exit__(*sys.exc_info())

c:\users\user\appdata\local\programs\python\python35\lib\contextlib.py in __enter__(self)
     57     def __enter__(self):
     58         try:
---> 59             return next(self.gen)
     60         except StopIteration:
     61             raise RuntimeError("generator didn't yield") from None

c:\users\user\appdata\local\programs\python\python35\lib\site-packages\tensorflow\python\framework\ops.py in name_scope(self, name)
   3942         # (viz. '-', '\', '/', and '_').
   3943         if not _VALID_SCOPE_NAME_REGEX.match(name):
-> 3944           raise ValueError("'%s' is not a valid scope name" % name)
   3945       else:
   3946         # Scopes created in the root must match the more restrictive

ValueError: 'Exterior1st_Wd Sdng' is not a valid scope name

I am not able to understand what the term "Exterior1st_Wd Sdng" means because I don't have any variable named like that. Thanks in advance!

Answer 1

I was trying to figure out exactly what scope names are allowed in Tensorflow and this page was the first result, so I'll take the opportunity to post the answer here for anyone else with the same query.

At the time of writing, allowable scope names appear to be restricted by this line in the Tensorflow source:

_VALID_SCOPE_NAME_REGEX = re.compile("^[A-Za-z0-9_.\\-/>]*$")

So in other words, scope names can contain letters (upper and lowercase), numbers, and _, ., \, -, /,>. (Notably, they can't include whitespace.)

Hope someone finds this helpful.

Answer 2

Probably the regexp _VALID_SCOPE_NAME_REGEX doesn't like the names of the dataset columns. Can you check the names of the columns in the housing_data pandas dataset? Do they contains symbols '-', '\', '/', or '_'? If yes, can you get rid of them and retry?

Answer 3

I am not sure this is related to your error, but I've never seen feature columns as a set, normally they are a list. I think the construct_feature_columns function should just return [tf.feature_column.numeric_column(my_feature) for my_feature in input_features]

Let me know if this works!!