Support for storing large tensor values in external files was introduced in #678, but AFAICT is undocumented.

This is a pretty important feature, functionally, but it's also important for end users who may not realise that they need to move around more than just the *.onnx file.

I would suggest it should be documented in IR.md, and perhaps there are other locations from which it could be s

I've updated TPOT to version '0.10.2', and if I run the following code:

tpot = TPOTClassifier(generations=200, population_size=200, scoring='precision',
                          template='wrong template')
tpot.fit(X_train, y_train)

I receive a RuntimeError:
`RuntimeError: There was an error in the TPOT optimization process. This could be because the data was not formatted

Suppose I have a dask dataframe with lots of columns, all of them of float dtype. _meta_nonempty creates each column separately and then makes a dataframe out of it. This makes some operations surprisingly slow, such as loc. I see some issues with this:

1. _meta_nonempty can be optimized for some dataframes.
2. loc shouldn't require _meta_nonempty in the first place. It only needs `s

Replace path_management fixture with pytest.tmpdir fixture

Bug/Feature Request Description:

Currently we define a custom fixture to clean up our serialization tests in entityset_tests.test_serialization. We can replace our custom fixture with pytest's tmpdir.

http://doc.pytest.org/en/latest/tmpdir.html

Hey all,

When running Auto-Sklearn in parallel with n_jobs and providing a True flag for delete_tmp_folder_after_terminate and delete_output_folder_after_terminate, the directories are not deleted in the end. This is not reflected in the documentation, where for both parameters the following is written:

delete_tmp_folder_after_terminate: string, optional (True)
remove tmp_folder,

sklearn allows to pass a dictionary as scorer for multimetric scoring. For skorch, a user should pass multiple scoring callbacks instead.

Right now, if a user passes a dict, skorch will fail while trying to retrieve a name here:

https://github.com/skorch-dev/skorch/blob/0581bc6062df86a46a7857ffeab0787e524abbe0/skorch/callbacks/scoring.py#L99

Or here if a name was passed:

https://github

After the visualizer audit we believe that the quick methods are ready to be made prime time.

Quick methods must:

1. Have a function signature identical to the Visualizer constructor with additional data for fit/score
2. Call finalize() and not poof()
3. Return the fitted visualizer instance
4. Some visualizers must also insure that they call transform/fit_transform/score

For eac

Follow an example where the change of this widget was already reported in the blog. Change the documentation accordingly.

I think it could be useful, when one wants to plot only e.g. class 1, to have an option to produce consistent plots for both plot_cumulative_gain and plot_roc

At the moment, instead, only plot_roc supports such option.

Thanks a lot

As listed in https://docs.scipy.org/doc/scipy/reference/special.html#bessel-functions, we can implement the universal functions in the list first.

I'm sorry if I missed this functionality, but CLI version hasn't it for sure (I saw the related code only in generate_code_examples.py). I guess it will be very useful to eliminate copy-paste phase, especially for large models.

Of course, piping is a solution, but not for development in Jupyter Notebook, for example.