dask xgboost AttributeError: 'DataFrame' object has no attribute 'unique'

[naarkhoo]

I am trying to do classification using dask xgboost
but get

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-60-d22c31ae208b> in <module>
      2 client = Client()
      3 
----> 4 est.fit(ttx, tty)

/opt/conda/lib/python3.8/site-packages/dask_xgboost/core.py in fit(self, X, y, classes, eval_set, sample_weight, sample_weight_eval_set, eval_metric, early_stopping_rounds)
    563                 classes = da.unique(y)
    564             else:
--> 565                 classes = y.unique()
    566             classes = classes.compute()
    567         else:

/opt/conda/lib/python3.8/site-packages/dask/dataframe/core.py in __getattr__(self, key)
   3912             return self[key]
   3913         else:
-> 3914             raise AttributeError("'DataFrame' object has no attribute %r" % key)
   3915 
   3916     def __dir__(self):

AttributeError: 'DataFrame' object has no attribute 'unique'

I am using dask '2021.04.1'; also I noticed a similar recent (25 days go) rapidsai/cuml#3381 - not sure if relevant. I don't have this problem if I use xgboost regressor.

[ncclementi]

@naarkhoo Can you please provide a minimal reproducible example? That way we can reproduce the error and try to figure out what's happening.

[pavithraes]

@naarkhoo Thanks for reporting! As @ncclementi suggested, a minimal example will allow us to help you better. :)

A few notes:

• dask-xgboost (which you seem to be using) is now deprecated, and we recommend using the direct Dask implementation in XGBoost.
• I'd suggest making sure "ttx" and "tty" are Dask Arrays or array-like data structures and not Dask DataFrames (or Series). If they are Dask DataFrames, you can use .to_dask_array to convert them to Arrays.

Also, this looks like more of a usage question than a bug report or feature request. We encourage people with such questions to ask them at the Dask Discourse. Would you mind opening a discussion topic about this over there? If you do so, feel free to close this issue and we will continue the conversation on discourse. :)

[naarkhoo]

that was the case - thanks.

…

On Mon, Jul 18, 2022 at 4:05 PM Pavithra Eswaramoorthy < ***@***.***> wrote: @naarkhoo <https://github.com/naarkhoo> Also, note that dask-xgboost <https://github.com/dask/dask-xgboost> (which you seem to be using) is now deprecated, and we recommend using the direct Dask implementation in XGBoost <https://xgboost.readthedocs.io/en/stable/tutorials/dask.html>. — Reply to this email directly, view it on GitHub <#9281 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABWWO34D7JETGBYW4ZAHJTVUVQA5ANCNFSM53W4DULQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>