1

通常的 ML 管道涉及将 pandas 或 dask 数据帧处理成可以传递给 ML 模型的形式。然而，许多 dask-ml 模型不能接受 Dask 数据帧，因为它们不跟踪每个分区的行数。调用该fit方法会抛出一个Cannot fit on dask.dataframe due to unknown partition lengths error. 我应该怎么做才能将 Dask 数据帧传递给 dask-ml 模型？

这是一个例子：

import dask.dataframe as dd
import pandas as pd
from dask_ml.cluster import KMeans

df = dd.from_pandas(pd.DataFrame({'A': [1, 2, 3, 4, 5], 
                                  'B': [6, 7, 8, 9, 10]}),
                    npartitions=2)

kmeans = KMeans()
kmeans.fit(df)


---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-53-6c1545864b12> in <module>()
      6 
      7 kmeans = KMeans()
----> 8 kmeans.fit(df)

~/anaconda3/envs/pds/lib/python3.6/site-packages/dask_ml/cluster/k_means.py in fit(self, X, y)
    187 
    188     def fit(self, X, y=None):
--> 189         X = self._check_array(X)
    190         labels, centroids, inertia, n_iter = k_means(
    191             X,

~/anaconda3/envs/pds/lib/python3.6/site-packages/dask_ml/utils.py in wraps(*args, **kwargs)
    298         def wraps(*args, **kwargs):
    299             with _timer(f.__name__, _logger=logger, level=level):
--> 300                 results = f(*args, **kwargs)
    301             return results
    302 

~/anaconda3/envs/pds/lib/python3.6/site-packages/dask_ml/cluster/k_means.py in _check_array(self, X)
    159         elif isinstance(X, dd.DataFrame):
    160             raise TypeError(
--> 161                 "Cannot fit on dask.dataframe due to unknown " "partition lengths."
    162             )
    163 

TypeError: Cannot fit on dask.dataframe due to unknown partition lengths.
4

1 回答 1

1

现在使用https://github.com/dask/dask-ml/pull/393支持 dask-ml master

这将包含在 Dask-ML 0.10 版本中。

于 2018-10-08T19:36:41.637 回答