1 files changed, 53 insertions, 21 deletions
diff --git a/tensorflow/contrib/boosted_trees/estimator_batch/model.py b/tensorflow/contrib/boosted_trees/estimator_batch/model.py
index 0e8a56e6e9..2fbe72951a 100644
--- a/tensorflow/contrib/boosted_trees/estimator_batch/model.py
+++ b/tensorflow/contrib/boosted_trees/estimator_batch/model.py
@@ -29,7 +29,17 @@ from tensorflow.python.framework import ops
 from tensorflow.python.ops import state_ops
 from tensorflow.python.training import training_util
 
-def model_builder(features, labels, mode, params, config):
+class ModelBuilderOutputType(object):
+  MODEL_FN_OPS = 0
+  ESTIMATOR_SPEC = 1
+
+
+def model_builder(features,
+                  labels,
+                  mode,
+                  params,
+                  config,
+                  output_type=ModelBuilderOutputType.MODEL_FN_OPS):
   """Multi-machine batch gradient descent tree model.
 
   Args:
@@ -115,31 +125,53 @@ def model_builder(features, labels, mode, params, config):
         return update_op
 
   create_estimator_spec_op = getattr(head, "create_estimator_spec", None)
-  if use_core_libs and callable(create_estimator_spec_op):
-    model_fn_ops = head.create_estimator_spec(
-        features=features,
-        mode=mode,
-        labels=labels,
-        train_op_fn=_train_op_fn,
-        logits=logits)
-    model_fn_ops = estimator_utils.estimator_spec_to_model_fn_ops(model_fn_ops)
-  else:
-    model_fn_ops = head.create_model_fn_ops(
-        features=features,
-        mode=mode,
-        labels=labels,
-        train_op_fn=_train_op_fn,
-        logits=logits)
-  if output_leaf_index and gbdt_batch.LEAF_INDEX in predictions_dict:
-    model_fn_ops.predictions[gbdt_batch.LEAF_INDEX] = predictions_dict[
-        gbdt_batch.LEAF_INDEX]
+
   if num_trees:
     if center_bias:
       num_trees += 1
     finalized_trees, attempted_trees = gbdt_model.get_number_of_trees_tensor()
-    model_fn_ops.training_hooks.append(
+    training_hooks = [
         trainer_hooks.StopAfterNTrees(num_trees, attempted_trees,
-                                      finalized_trees))
+                                      finalized_trees)
+    ]
+
+  if output_type == ModelBuilderOutputType.MODEL_FN_OPS:
+    if use_core_libs and callable(create_estimator_spec_op):
+      model_fn_ops = head.create_estimator_spec(
+          features=features,
+          mode=mode,
+          labels=labels,
+          train_op_fn=_train_op_fn,
+          logits=logits)
+      model_fn_ops = estimator_utils.estimator_spec_to_model_fn_ops(
+          model_fn_ops)
+    else:
+      model_fn_ops = head.create_model_fn_ops(
+          features=features,
+          mode=mode,
+          labels=labels,
+          train_op_fn=_train_op_fn,
+          logits=logits)
+
+    if output_leaf_index and gbdt_batch.LEAF_INDEX in predictions_dict:
+      model_fn_ops.predictions[gbdt_batch.LEAF_INDEX] = predictions_dict[
+          gbdt_batch.LEAF_INDEX]
+
+    model_fn_ops.training_hooks.extend(training_hooks)
+    return model_fn_ops
+  elif output_type == ModelBuilderOutputType.ESTIMATOR_SPEC:
+    assert callable(create_estimator_spec_op)
+    estimator_spec = head.create_estimator_spec(
+        features=features,
+        mode=mode,
+        labels=labels,
+        train_op_fn=_train_op_fn,
+        logits=logits)
+
+    estimator_spec = estimator_spec._replace(
+        training_hooks=training_hooks + list(estimator_spec.training_hooks))
+    return estimator_spec
+
   return model_fn_ops