一個numpy int64的坑 頂 原

今天跑一個tensorflow代碼，由於最後結果需要導出爲json的文件，所以採用了json.dumps直接寫文件，但是一直報錯：

Traceback (most recent call last):
  File "predict_test.py", line 62, in <module>
    json.dump(submit, fw)
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/__init__.py", line 178, in dump
    for chunk in iterable:
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 427, in _iterencode
    yield from _iterencode_list(o, _current_indent_level)
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 324, in _iterencode_list
    yield from chunks
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 403, in _iterencode_dict
    yield from chunks
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 324, in _iterencode_list
    yield from chunks
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 436, in _iterencode
    o = _default(o)
  File "/home/tensorflow/.pyenv/versions/3.5.3/lib/python3.5/json/encoder.py", line 179, in default
    raise TypeError(repr(o) + " is not JSON serializable")
TypeError: 26 is not JSON serializable

矇蔽了很久，怎麼26這個數字不可以serializable？？

定位到報錯代碼位置，報錯代碼如下：

top_indices = []
submit = []
for i in range(n):
    diction = {}
    diction["label_id"] = list(preds[i].argsort()[-3:][::-1])
    diction["image_id"] = id_list[i]
    submit.append(diction)

with open('submit.plk','w') as fw:
    json.dump(submit, fw)

好像沒有什麼問題呀~？？

做了個小測試：

In [2]: a = submit[1]

In [3]: a
Out[3]: 
{'image_id': 'c532ed79f6c54fb3cf8c45004656e385079a5284.jpg',
 'label_id': [25, 13, 15]}

In [4]: b = {'image_id': 'c532ed79f6c54fb3cf8c45004656e385079a5284.jpg',
   ...:  'label_id': [25, 13, 15]}

In [5]: json.dumps(b)
Out[5]: '{"image_id": "c532ed79f6c54fb3cf8c45004656e385079a5284.jpg", "label_id": [25, 13, 15]}'

In [6]: a == b
Out[6]: True

In [7]: json.dumps(a)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-9-373a2a7edfd2> in <module>()
...
TypeError: 25 is not JSON serializable

明明a 的值 和b 相等，爲什麼b就可以json.dumps，而 a json.dumps報錯，懵逼一會，想想要不查看下25這個對象，原來對於a來說 25 是一個<type 'numpy.int64'>！！！

被狠狠地坑了半個小時~~由於平時做數據處理大家會經常用到numpy的array，所以大家也要注意，雖然都是整形，但numpy.int64和python的整數是不一樣的，解決方案，可以用astype('in32t')或者直接用python的int()方法。