No need for OrderedDict, default dictionary has been in order for a while

https://realpython.com/python-ordereddict/ for more info or look at Python changelings.

src/python/twitter/deepbird/projects/timelines/scripts/models/earlybird/metrics.py

@@ -1,6 +1,5 @@
 # checkstyle: noqa
 import tensorflow.compat.v1 as tf
-from collections import OrderedDict
 from .constants import EB_SCORE_IDX
 from .lolly.data_helpers import get_lolly_scores
 
@@ -35,7 +34,7 @@ def get_multi_binary_class_metric_fn(metrics, classes=None, class_dim=1):
     # Added to support per engagement metrics for both TF and Lolly scores.
     labels = tf.tile(labels, [1, 2])
 
-    eval_metric_ops = OrderedDict()
+    eval_metric_ops = dict()
 
     preds = graph_output['output']