Fix for pandas-dev#16909

get_loc should now work for np.timedelta64 data type.

doc/source/whatsnew/v0.21.0.txt

@@ -155,6 +155,7 @@ Indexing
 - When called with a null slice (e.g. ``df.iloc[:]``), the ``.iloc`` and ``.loc`` indexers return a shallow copy of the original object. Previously they returned the original object. (:issue:`13873`).
 - When called on an unsorted ``MultiIndex``, the ``loc`` indexer now will raise ``UnsortedIndexError`` only if proper slicing is used on non-sorted levels (:issue:`16734`).
 - Fixes regression in 0.20.3 when indexing with a string on a ``TimedeltaIndex`` (:issue:`16896`).
+- Fixed ``TimedeltaIndex.get_loc`` ``for np.timedelta64`` data type(:issue:`16909`).
 
 I/O
 ^^^

pandas/core/indexes/timedeltas.py

@@ -682,7 +682,7 @@ def get_loc(self, key, method=None, tolerance=None):
         -------
         loc : int
         """
-        if is_bool_indexer(key) or is_timedelta64_dtype(key):
+        if is_list_like(key):
             raise TypeError
 
         if isnull(key):

pandas/tests/indexes/timedeltas/test_timedelta.py

@@ -66,6 +66,8 @@ def test_get_loc(self):
         for method, loc in [('pad', 1), ('backfill', 2), ('nearest', 1)]:
             assert idx.get_loc('1 day 1 hour', method) == loc
 
+        assert idx.get_loc(idx[1].to_timedelta64()) == 1
+
         # GH 16896
         assert idx.get_loc('0 days') == 0