tensorflow - 在 Colab 中执行联邦学习文本生成教程时出错

Question

我正在尝试遵循有关联邦学习 TensorFlow 的本教程，并且在执行此行时出现错误：

train_data, test_data = tff.simulation.datasets.shakespeare.load_data()

错误：

    Downloading shakespeare.sqlite.lzma:  79%|███████▉  | 1048576/1329828 [00:00<00:00, 12187174.26it/s]
---------------------------------------------------------------------------
NotFoundError                             Traceback (most recent call last)
<ipython-input-12-f8f1fc62c096> in <module>()
----> 1 train_data, test_data = tff.simulation.datasets.shakespeare.load_data()

4 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/lib/io/file_io.py in _prewrite_check(self)
     86                                            "File isn't open for writing")
     87       self._writable_file = _pywrap_file_io.WritableFile(
---> 88           compat.path_to_bytes(self.__name), compat.as_bytes(self.__mode))
     89 
     90   def _prepare_value(self, val):

NotFoundError: /root/.tff/shakespeare.sqlite; No such file or directory

在此处输入图像描述

Answer 1

这是当前tensorflow-federated-nightlypip 包的问题。它应该由https://github.com/tensorflow/federated/pull/1162修复。

同时，尝试将pip install第一个单元格中的更改为使用tensorflow-federated包而不是 nightly 包。例子：

!pip install --quiet --upgrade tensorflow_federated
!pip install --quiet --upgrade nest_asyncio

import nest_asyncio
nest_asyncio.apply()