Remove contrib cross device ops and update all_reduce_alg options. (#6673)

* Remove contrib AllReduceCrossDeviceOps and update all_reduce_alg options with MirroredStrategy.

* cleanup

official/utils/misc/distribution_utils.py

@@ -30,6 +30,16 @@ _COLLECTIVE_COMMUNICATION_OPTIONS = {
     "nccl": tf.distribute.experimental.CollectiveCommunication.NCCL
 }
 
+_MIRRORED_ALL_REDUCE_NUM_PACKS = 2
+
+_MIRRORED_ALL_REDUCE_OPTIONS = {
+    None: None,
+    "nccl": tf.distribute.NcclAllReduce(
+        num_packs=_MIRRORED_ALL_REDUCE_NUM_PACKS),
+    "hierarchical_copy": tf.distribute.HierarchicalCopyAllReduce(
+        num_packs=_MIRRORED_ALL_REDUCE_NUM_PACKS)
+}
+
 
 def get_distribution_strategy(distribution_strategy="default",
                               num_gpus=0,
@@ -38,20 +48,19 @@ def get_distribution_strategy(distribution_strategy="default",
   """Return a DistributionStrategy for running the model.
 
   Args:
-    distribution_strategy: a string specify which distribution strategy to use.
-      Accepted values are 'off', 'default', 'one_device', 'mirrored',
+    distribution_strategy: a string specifying which distribution strategy to
+      use. Accepted values are 'off', 'default', 'one_device', 'mirrored',
       'parameter_server', 'multi_worker_mirrored', case insensitive. 'off' means
       not to use Distribution Strategy; 'default' means to choose from
       `MirroredStrategy`, `MultiWorkerMirroredStrategy`, or `OneDeviceStrategy`
       according to the number of GPUs and number of workers.
     num_gpus: Number of GPUs to run this model.
     num_workers: Number of workers to run this model.
-    all_reduce_alg: Optional. Specify which algorithm to use when performing
-      all-reduce. See tf.contrib.distribute.AllReduceCrossDeviceOps for
-      available algorithms when used with `mirrored`, and
-      tf.distribute.experimental.CollectiveCommunication when used with
-      `multi_worker_mirrored`. If None, DistributionStrategy will choose based
-      on device topology.
+    all_reduce_alg: Optional. Specifies which algorithm to use when performing
+      all-reduce. For `MirroredStrategy`, valid values are "nccl" and
+      "hierarchical_copy". For `MultiWorkerMirroredStrategy`, valid values are
+      "ring" and "nccl".  If None, DistributionStrategy will choose based on
+      device topology.
 
   Returns:
     tf.distribute.DistibutionStrategy object.
@@ -74,7 +83,7 @@ def get_distribution_strategy(distribution_strategy="default",
     if all_reduce_alg not in _COLLECTIVE_COMMUNICATION_OPTIONS:
       raise ValueError(
           "When used with `multi_worker_mirrored`, valid values for "
-          "all_reduce_alg are [`ring`, `nccl`].  Supplied value: {}".format(
+          "all_reduce_alg are ['ring', 'nccl'].  Supplied value: {}".format(
               all_reduce_alg))
     return tf.distribute.experimental.MultiWorkerMirroredStrategy(
         communication=_COLLECTIVE_COMMUNICATION_OPTIONS[all_reduce_alg])
@@ -95,13 +104,14 @@ def get_distribution_strategy(distribution_strategy="default",
       devices = ["device:CPU:0"]
     else:
       devices = ["device:GPU:%d" % i for i in range(num_gpus)]
-    if all_reduce_alg:
-      return tf.distribute.MirroredStrategy(
-          devices=devices,
-          cross_device_ops=tf.contrib.distribute.AllReduceCrossDeviceOps(
-              all_reduce_alg, num_packs=2))
-    else:
-      return tf.distribute.MirroredStrategy(devices=devices)
+    if all_reduce_alg not in _MIRRORED_ALL_REDUCE_OPTIONS:
+      raise ValueError(
+          "When used with `mirrored`, valid values for all_reduce_alg are "
+          "['nccl', 'hierarchical_copy'].  Supplied value: {}".format(
+              all_reduce_alg))
+    return tf.distribute.MirroredStrategy(
+        devices=devices,
+        cross_device_ops=_MIRRORED_ALL_REDUCE_OPTIONS[all_reduce_alg])
 
   if distribution_strategy == "parameter_server":
     return tf.distribute.experimental.ParameterServerStrategy()