Aaron Richter

Pickle5 error with scikit-learn loky backend on Linux with Python<3.8

1

Hello! I was getting the following error when training a scikit-learn model with n_jobs=-1, and I see that its been mentioned before on different issues in related packages [here](https://github.com/scikit-learn-contrib/skope-rules/issues/18) and...

Multi-class support for LogisticRegression

3

(splitting out a request from #386 ) LogisticRegression currently only supports binary classification (the multi_class argument is ignored). This feature request is to add multi-class support!

ParallelPostFit excessive scheduler memory use and CancelledError

1

Notebook with MCVE and all notes: https://nbviewer.jupyter.org/gist/rikturr/43336377678018d01d4f21f19dd7ef11 When using `ParallelPostFit` to train with pandas/numpy objects then predict on dask objects, I noticed that the scheduler memory use runs extremely high....

map_partitions or map_blocks with large objects eats up scheduler memory

5

Opening this report from investigation in https://github.com/dask/dask-ml/issues/842 and https://github.com/dask/dask-ml/pull/843 When passing some large objects as arguments to map_blocks / map_partitions, the scheduler memory can quickly be overwhelmed. Large objects _should_...