DataIterator deterministically picks the 1st dataset for creating TF …

…iterators. Fixing #23
asyml · Sep 14, 2018 · 03ad846 · 03ad846
1 parent 1fa6b4d
commit 03ad846
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 6 deletions.
diff --git a/texar/data/data/data_base.py b/texar/data/data/data_base.py
@@ -78,7 +78,10 @@ def default_hparams():
             "allow_smaller_final_batch" : bool
                Whether to allow the final batch to be smaller if there are
                insufficient elements left. If `False`, the final batch is
-               discarded if it is smaller than batch size.
+               discarded if it is smaller than batch size. Note that,
+               if `True`, `output_shapes` of the resulting dataset
+               will have a a **static** batch_size dimension equal to
+               "batch_size".
 
             "shuffle" : bool
                 Whether to randomly shuffle the elements of the dataset.

diff --git a/texar/data/data/data_iterators.py b/texar/data/data/data_iterators.py
@@ -135,9 +135,9 @@ def __init__(self, datasets):
 
         self._variable_scope = get_unique_named_variable_scope('data_iterator')
         with tf.variable_scope(self._variable_scope):
-            arb_dataset = self._datasets[next(iter(self._datasets))]
+            first_dataset = self._datasets[sorted(self.dataset_names)[0]]
             self._iterator = tf.data.Iterator.from_structure(
-                arb_dataset.output_types, arb_dataset.output_shapes)
+                first_dataset.output_types, first_dataset.output_shapes)
             self._iterator_init_ops = {
                 name: self._iterator.make_initializer(d)
                 for name, d in self._datasets.items()
@@ -324,10 +324,10 @@ def __init__(self, datasets):
             'feedable_data_iterator')
         with tf.variable_scope(self._variable_scope):
             self._handle = tf.placeholder(tf.string, shape=[], name='handle')
-            arb_dataset = self._datasets[next(iter(self._datasets))]
+            first_dataset = self._datasets[sorted(self.dataset_names)[0]]
             self._iterator = tf.data.Iterator.from_string_handle(
-                self._handle, arb_dataset.output_types,
-                arb_dataset.output_shapes)
+                self._handle, first_dataset.output_types,
+                first_dataset.output_shapes)
 
             self._dataset_iterators = {
                 name: dataset.make_initializable_iterator()