AttributeError:模块“tensorflow_federated.python.simulation”没有属性“HDF5ClientData”

Question

我有一个数据集写到一个totff.simulation.datasets.ClientData，文件中，我想要将它转换为上的表单h5。一周前，我用这个读到了

train_path='FederatedClients/dataTrain.h5'
train_data = tff.simulation.HDF5ClientData(train_path)

但是现在，我再次得到了这个语句的以下错误。

----------------------------------------------------------
AttributeError           Traceback (most recent call last)
<ipython-input-6-f4c34c62ca50> in <module>
      1 train_path='FederatedClients/dataTrain.h5'
      2 #train_data=pd.read_hdf(train_path)
----> 3 train_data = tff.simulation.HDF5ClientData(train_path)
      4 test_path='FederatedClients/dataTest.h5'
      5 test_data=pd.read_hdf(test_path)

AttributeError: module 'tensorflow_federated.python.simulation' has no attribute 'HDF5ClientData'

我不知道该怎么做了，我用最基本的熊猫来读取H5文件，我也得到了错误，TAT

train_data=pd.read_hdf(train_path)

----------------------------------------------------------
ValueError               Traceback (most recent call last)
<ipython-input-7-151fe098668c> in <module>
      1 train_path='FederatedClients/dataTrain.h5'
----> 2 train_data=pd.read_hdf(train_path)
      3 train_data = tff.simulation.HDF5ClientData(train_path)
      4 test_path='FederatedClients/dataTest.h5'
      5 test_data=pd.read_hdf(test_path)

~/anaconda3/envs/tff/lib/python3.7/site-packages/pandas/io/pytables.py in read_hdf(path_or_buf, key, mode, errors, where, start, stop, columns, iterator, chunksize, **kwargs)
    437             if len(groups) == 0:
    438                 raise ValueError(
--> 439                     "Dataset(s) incompatible with Pandas data types, "
    440                     "not table, or no datasets found in HDF5 file."
    441                 )

ValueError: Dataset(s) incompatible with Pandas data types, not table, or no datasets found in HDF5 file.

如果你能解决我的问题，我将不胜感激。

Answer 1

TFF版本0.19.0删除了tff.simulation.HDF5ClientData (发布说明)。TFF数据集现在由SQLite而不是HDF5支持。

HDF5实现可以从data.py中恢复。在本地复制并替换：

• client_data.ClientData和tff.simulation.datasets.ClientData。
• 删除py_typecheck和tensor_utils的使用，它们是不必要的。