KMeans fit TypeError expected dtype object, got 'numpy.dtype[float64]'

What happened: when fit X, get error:

TypeError                                 Traceback (most recent call last)
<ipython-input-11-7d3bc4c475f7> in <module>
      1 km = dask_ml.cluster.KMeans(n_clusters=3, init_max_iter=2, oversampling_factor=10)
----> 2 km.fit(X)
 ...
 ...
 TypeError: expected dtype object, got 'numpy.dtype[float64]'

What you expected to happen: output: KMeans(init_max_iter=2, n_clusters=3, oversamplinh_factor=10)

Minimal Complete Verifiable Example:

import dask_ml.cluster

X, y = dask_ml.datasets.make_blobs(n_samples=1000000,
                                   chunks=100000,
                                   random_state=0,
                                   centers=3)
km = dask_ml.cluster.KMeans(n_clusters=3, init_max_iter=2, oversampling_factor=10)
km.fit(X)

Environment:

• Dask version: dask_ml 1.8.0, dask 2021.02.0
• Python version: 3.7
• Operating System: MacOS Catalina 10.15.5
• Install method (conda, pip, source): conda
• Other packages: numpy 1.20.1, pandas 1.2.2

[hristog]

Thanks for the further info provided, @Mimran0715!

I confirm I’ve been able to reproduce the issue locally, using the following versions:

• numpy==1.20.1
• pandas==0.25.1
• numba==0.50.0 (was running into https://github.com/mkocabas/VIBE/issues/88 and https://github.com/numba/numba/issues/5415 with 0.45.1)
• llvmlite==0.32.1
• dask==2.5.2
• dask-glm==0.2.0
• dask-ml==1.8.0
• Python 3.7.3

As suggested in my comment above, this was resolved via upgrading numba to the latest available version. I just did python3 -m pip install numba==0.53.0, and this has enabled me to run the K-Means|| example.

Please, have a go at upgrading the numba version (you might need to update llvmlite to something compatible as well, e.g., 0.32.1; if you do, please, take a note of your current version of llvmlite, so that you can always revert back later on, if needs be). If you don’t want to globally affect your local versions, you can test with an updated version of numba via venv or via a Docker container.

Thanks!

[Mimran0715]

Thanks for responding. I’m just trying to run the K-means cluster example here: https://examples.dask.org/machine-learning/training-on-large-datasets.html Here is the Traceback: Here are some details abt the package versions: