使用keras.utils.get_file()触发TypeError：int与NoneType无法比较

环境信息 #

• Windows 10
• Python 3.9.19
• Tensorflow 2.10.1

代码片段 #

dataset_path = keras.utils.get_file("auto-mpg.data", "https://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data")

问题描述 #

学习《Applied Deep Learning and Computer Vision for Self-Driving Cars》一书时，运行示例代码调用keras.utils.get_file()获取数据集文件，触发TypeError，错误提示'<' not supported between instances of 'int' and 'NoneType'。期望成功下载并获取文件路径，已查阅相关Stack Overflow帖子但仅得到“It works for me”的回复，无法解决问题。

报错详情 #

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[2], line 2
      1 # Datapath to import auto-mpg data
----> 2 dataset_path = keras.utils.get_file("auto-mpg.data", "https://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data")

File ~\anaconda3\envs\AVTech\lib\site-packages\keras\utils\data_utils.py:296, in get_file(fname, origin, untar, md5_hash, file_hash, cache_subdir, hash_algorithm, extract, archive_format, cache_dir)
    294 try:
    295     try:
---> 296         urlretrieve(origin, fpath, DLProgbar())
    297     except urllib.error.HTTPError as e:
    298         raise Exception(error_msg.format(origin, e.code, e.msg))

File ~\anaconda3\envs\AVTech\lib\site-packages\keras\utils\data_utils.py:86, in urlretrieve(url, filename, reporthook, data)
     84 response = urlopen(url, data)
     85 with open(filename, "wb") as fd:
---> 86     for chunk in chunk_read(response, reporthook=reporthook):
     87         fd.write(chunk)

File ~\anaconda3\envs\AVTech\lib\site-packages\keras\utils\data_utils.py:78, in urlretrieve.<locals>.chunk_read(response, chunk_size, reporthook)
     76 count += 1
     77 if reporthook is not None:
---> 78     reporthook(count, chunk_size, total_size)
     79 if chunk:
     80     yield chunk

File ~\anaconda3\envs\AVTech\lib\site-packages\keras\utils\data_utils.py:287, in get_file.<locals>.DLProgbar.__call__(self, block_num, block_size, total_size)
    285     self.progbar = Progbar(total_size)
    286 current = block_num * block_size
---> 287 if current < total_size:
    288     self.progbar.update(current)
    289 elif not self.finished:

TypeError: '<' not supported between instances of 'int' and 'NoneType'

解决方案 #

该错误源于服务器返回的响应未提供文件总大小（Content-Length），导致total_size为None，进度条更新时触发类型不匹配错误，可通过以下方式解决：

• 跳过进度条逻辑：
  修改get_file调用，手动设置reporthook=None，绕过进度条的更新判断：

  dataset_path = keras.utils.get_file(
      "auto-mpg.data",
      "https://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data",
      reporthook=None
  )
  

• 手动指定本地文件路径：
  直接访问URL下载auto-mpg.data文件，保存到本地后，将文件路径直接赋值给变量：

  dataset_path = "本地文件存储路径/auto-mpg.data"
  

• 升级TensorFlow版本：
  该问题在TensorFlow 2.11及以上版本中已被修复，升级到稳定新版本即可解决：

  pip install --upgrade tensorflow
  

内容的提问来源于stack exchange，提问作者RPBruiser