Internal change

PiperOrigin-RevId: 352730170

official/core/input_reader.py

@@ -30,6 +30,10 @@ def _get_random_integer():
 class InputReader:
   """Input reader that returns a tf.data.Dataset instance."""
 
+  # A static random number which is the same across different InputReader
+  # instances.
+  static_randnum = _get_random_integer()
+
   def __init__(self,
                params: cfg.DataConfig,
                dataset_fn=tf.data.TFRecordDataset,
@@ -136,7 +140,13 @@ class InputReader:
     self._enable_tf_data_service = (
         params.enable_tf_data_service and params.tf_data_service_address)
     self._tf_data_service_address = params.tf_data_service_address
-    self._tf_data_service_job_name = params.tf_data_service_job_name
+    if self._enable_tf_data_service:
+      # Add a random seed as the tf.data service job name suffix, so tf.data
+      # service doesn't reuse the previous state if TPU worker gets preempted.
+      self._tf_data_service_job_name = (
+          params.tf_data_service_job_name + str(self.static_randnum))
+      self._enable_round_robin_tf_data_service = params.get(
+          'enable_round_robin_tf_data_service', False)
 
   def _shard_files_then_read(
       self, input_context: Optional[tf.distribute.InputContext] = None):
@@ -276,7 +286,30 @@ class InputReader:
 
     dataset = maybe_map_fn(dataset, self._postprocess_fn)
 
-    if self._enable_tf_data_service:
+    if self._enable_tf_data_service and input_context:
+      if self._enable_round_robin_tf_data_service:
+        replicas_per_input_pipeline = input_context.num_replicas_in_sync // (
+            input_context.num_input_pipelines)
+        base_consumer_index = input_context.input_pipeline_id * (
+            replicas_per_input_pipeline)
+        num_consumers = input_context.num_input_pipelines * (
+            replicas_per_input_pipeline)
+        range_dataset = tf.data.Dataset.range(replicas_per_input_pipeline)
+        dataset = range_dataset.map(lambda i: dataset.apply(  # pylint: disable=g-long-lambda
+            tf.data.experimental.service.distribute(
+                processing_mode='parallel_epochs',
+                service=self._tf_data_service_address,
+                job_name=self._tf_data_service_job_name,
+                consumer_index=base_consumer_index + i,
+                num_consumers=num_consumers)))
+        # Use parallel interleave to read multiple batches from a tf.data
+        # service worker in parallel.
+        dataset = dataset.interleave(
+            lambda x: x,
+            cycle_length=replicas_per_input_pipeline,
+            num_parallel_calls=replicas_per_input_pipeline,
+            deterministic=True)
+      else:
         dataset = dataset.apply(
             tf.data.experimental.service.distribute(
                 processing_mode='parallel_epochs',