Bug fix on flag name in wide_deep.py (#3777)

PR #3460 renamed FLAGS.epochs_per_eval to flags.epochs_between_evals; it missed this one

Author: nkconnor

official/wide_deep/wide_deep.py

@@ -182,9 +182,9 @@ def main(argv):
   train_file = os.path.join(flags.data_dir, 'adult.data')
   test_file = os.path.join(flags.data_dir, 'adult.test')
 
-  # Train and evaluate the model every `FLAGS.epochs_per_eval` epochs.
+  # Train and evaluate the model every `flags.epochs_between_evals` epochs.
   def train_input_fn():
-    return input_fn(train_file, flags.epochs_per_eval, True, flags.batch_size)
+    return input_fn(train_file, flags.epochs_between_evals, True, flags.batch_size)
 
   def eval_input_fn():
     return input_fn(test_file, 1, False, flags.batch_size)
@@ -194,7 +194,7 @@ def eval_input_fn():
       tensors_to_log={'average_loss': 'head/truediv',
                       'loss': 'head/weighted_loss/Sum'})
 
-  # Train and evaluate the model every `FLAGS.epochs_between_evals` epochs.
+  # Train and evaluate the model every `flags.epochs_between_evals` epochs.
   for n in range(flags.train_epochs // flags.epochs_between_evals):
     model.train(input_fn=train_input_fn, hooks=train_hooks)
     results = model.evaluate(input_fn=eval_input_fn)