AttributeError: module 'cupy' has no attribute 'array_equal'

[pseudotensor]

2020-12-17 16:40:24,010 C: 16% D:51.4GB  M:47.6GB  NODE:LOCAL1      25343  INFO   |     return f(**kwargs)
2020-12-17 16:40:24,011 C: 16% D:51.4GB  M:47.6GB  NODE:LOCAL1      25343  INFO   |   File "/home/jon/minicondadai/lib/python3.6/site-packages/xgboost/sklearn.py", line 845, in fit
2020-12-17 16:40:24,011 C: 16% D:51.4GB  M:47.6GB  NODE:LOCAL1      25343  INFO   |     if not cp.array_equal(self.classes_, cp.arange(self.n_classes_)):
2020-12-17 16:40:24,011 C: 16% D:51.4GB  M:47.6GB  NODE:LOCAL1      25343  INFO   | AttributeError: module 'cupy' has no attribute 'array_equal'

I guess 1.3.0 updated use of cupy API, but I don't see that xgboost has a cupy version dependency.

But cupy==7.8.0 does not work with this change.

xgboost only says need cupy-cuda101 for win64_test.yml and version 7.8.0 is fine version.

Did I miss some dependency somewhere else?

[pseudotensor]

@hcho3 from your PR commit c8ec621 related to label encoding.

[hcho3]

@pseudotensor My bad, cupy.array_equal was added in cuPy 8.0.0. I will submit a pull request to remove it.

[hcho3]

We will release a patch release with #6528. Thanks!