Automated rollback of change 136502135

Change: 136641403

tensorflow/contrib/learn/python/learn/estimators/estimator.py

@@ -216,12 +216,8 @@ def _make_metrics_ops(metrics, features, targets, predictions):
       predictions.
 
   Returns:
-    `dict` whose keys are summary names, and values are the result of the
-    metric, either:
-      - `Tensor` values (in which case only the result of the last eval batch
-        will be summarized).
-      - `tuple` of 2 `Tensor` objects, update op and value. The update op will
-        be run once each eval step, and the value written to summary.
+    A dict mapping the friendly given in `metrics` to the result of calling the
+    given metric function.
 
   Raises:
     ValueError: If metrics specifications do not work with the type of
@@ -271,13 +267,6 @@ def _make_metrics_ops(metrics, features, targets, predictions):
   return result
 
 
-def _maybe_add_streaming_mean(result, key, value):
-  if key in result:
-    logging.warning('Metrics already contains %s, skipping.', key)
-    return
-  result[key] = metrics_lib.streaming_mean(value)
-
-
 class BaseEstimator(
     sklearn.BaseEstimator, evaluable.Evaluable, trainable.Trainable):
   """Abstract BaseEstimator class to train and evaluate TensorFlow models.
@@ -585,7 +574,7 @@ class BaseEstimator(
     Args:
       features: `Tensor` or `dict` of `Tensor` objects.
       targets: `Tensor` or `dict` of `Tensor` objects.
-      metrics: Dict of metrics to run. If `None`, the default metric functions
+      metrics: Dict of metrics to run. If None, the default metric functions
         are used; if {}, no metrics are used. Otherwise, `metrics` should map
         friendly names for the metric to a `MetricSpec` object defining which
         model outputs to evaluate against which targets with which metric
@@ -1055,28 +1044,14 @@ class Estimator(BaseEstimator):
         `../metric_spec.py`.
 
     Returns:
-      `dict` whose keys are summary names, and values are either:
-        - `Tensor` values (in which case only the result of the last eval batch
-          will be summarized).
-        - `tuple` of 2 `Tensor` objects, update op and value. The update op will
-          be run once each eval step, and the value written to summary.
+      metrics: `dict` of `Tensor` objects.
 
     Raises:
       ValueError: if `metrics` don't match `targets`.
     """
     predictions, loss, _ = self._call_model_fn(features, targets, ModeKeys.EVAL)
-    result = _make_metrics_ops(metrics, features, targets, predictions)
-    _maybe_add_streaming_mean(result, 'loss', loss)
-
-    # TODO(ptucker): Work-around until we have an easier way to specify metrics
-    # from model_fn.
-    if predictions is not None:
-      if isinstance(predictions, dict):
-        for k, v in six.iteritems(predictions):
-          _maybe_add_streaming_mean(result, k, v)
-      else:
-        _maybe_add_streaming_mean(result, 'predictions', predictions)
-
+    result = {'loss': metrics_lib.streaming_mean(loss)}
+    result.update(_make_metrics_ops(metrics, features, targets, predictions))
     return result
 
   def _get_predict_ops(self, features):

tensorflow/contrib/learn/python/learn/graph_actions.py

@@ -25,7 +25,8 @@ import threading
 import time
 
 import numpy as np
-import six
+
+from six import reraise
 
 from tensorflow.contrib.framework import load_variable
 from tensorflow.contrib.framework.python.ops import ops as contrib_ops
@@ -578,7 +579,7 @@ def _train_internal(graph,
         logging.error('Got exception during tf.learn final checkpoint %s.', e)
       finally:
         if excinfo:
-          six.reraise(*excinfo)
+          reraise(*excinfo)
     return loss_value
 
 
@@ -628,14 +629,14 @@ def _write_summary_results(output_dir, eval_results, current_global_step):
                _eval_results_to_str(eval_results))
   summary_writer = get_summary_writer(output_dir)
   summary = summary_pb2.Summary()
-  for key, eval_result in six.iteritems(eval_results):
+  for key in eval_results:
     if eval_results[key] is None:
       continue
     value = summary.value.add()
     value.tag = key
-    if (isinstance(eval_result, np.float32) or
-        isinstance(eval_result, float)):
-      value.simple_value = float(eval_result)
+    if (isinstance(eval_results[key], np.float32) or
+        isinstance(eval_results[key], float)):
+      value.simple_value = float(eval_results[key])
     else:
       logging.warn('Skipping summary for %s, must be a float or np.float32.',
                    key)