Find keys in multidimensional dict based on value

I have a dict of dicts that looks like the following:

{ 'a':{'foo': True,
       'bar': 1,
       'baz': 'abc'},开发者_运维技巧
  'b':{'foo': True,
       'bar': 21,
       'baz': 'abc'},
  'c':{'foo': True,
       'bar': 3,
       'baz': 'cba'}}

What is the most efficient way to filter the dict to only include those keys whose sub-values for 'baz' are 'abc'? So in the example above, it would return a dict that looked like:

{ 'a':{'foo': True,
       'bar': 1,
       'baz': 'abc'},
  'b':{'foo': True,
       'bar': 21,
       'baz': 'abc'}}

---

{k: v for k, v in my_dict.items() if v['baz'] == 'abc'}