TypeError: a float is Required in sklearn.feature_extraction.FeatureHasher (version 0.16.1)

Hi, 
   I have noticed a bug in FeatureHasher (or maybe i am using it wrong).

Here is a slmple example to reproduce the bug: 

``` python
data = [{"device_id": "device1", "ad_id":"ad1"}, {"ad_id": "ad2", "other_feature": 5}]
FeatureHasher().fit_transform(data)
```
