AttributeError: module 'umap.plot' has no attribute 'output_notebook'

import umap
import umap.plot
umap.__version__

'0.4.2'.

After import umap in Jupyter notebook I get the same issue as in https://github.com/lmcinnes/umap/issues/321. I assumed that with version >0.4 it should have been resolved. Am I doing something wrong?

umap.plot.output_notebook()

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-2-c970ec05c096> in <module>
----> 1 umap.plot.output_notebook()

AttributeError: module 'umap.plot' has no attribute 'output_notebook'

And I indeed do not see 'output_notebook'

dir(umap.plot)

['Patch',
 '__builtins__',
 '__cached__',
 '__doc__',
 '__file__',
 '__loader__',
 '__name__',
 '__package__',
 '__spec__',
 '_blue',
 '_datashade_points',
 '_diagnostic_types',
 '_embed_datashader_in_an_axis',
 '_get_extent',
 '_green',
 '_matplotlib_points',
 '_nhood_compare',
 '_nhood_search',
 '_red',
 '_select_font_color',
 '_themes',
 '_to_hex',
 'bd',
 'bpl',
 'btr',
 'colorcet',
 'connectivity',
 'darkblue_cmap',
 'darkgreen_cmap',
 'darkpurple_cmap',
 'darkred_cmap',
 'deheap_sort',
 'diagnostic',
 'ds',
 'fire_cmap',
 'hd',
 'hv',
 'initialise_search',
 'initialized_nnd_search',
 'interactive',
 'matplotlib',
 'np',
 'numba',
 'pd',
 'plt',
 'points',
 'show',
 'show_interactive',
 'show_static',
 'sklearn',
 'submatrix',
 'tf',
 'warn']

As a workaround I use the function directly from bokeh

from bokeh.plotting import show, save, output_notebook, output_file

p = umap.plot.interactive(mapper, labels=labels, hover_data=hover_data)

And then I have to run the show command twice

output_notebook()
show(p)

In order not to get this error

RuntimeError: Models must be owned by only a single document, Circle(id='1036', ...) is already in a doc

Start by reproducing the issue in a Jupyter notebook with umap.plot.interactive and the Bokeh output_notebook workaround described in the report. Read the umap.plot interactive path and check how notebook output is initialized; done means output_notebook is available or clearly documented, and displaying the plot no longer requires repeated show calls or triggers the duplicate-document error.