add fit and predict number of samples to wish list

[garland-culbreth]

Changes

wish_list.xrst

• Add new item wishing for a way to specify the number of samples used in fit separately from the number used in predicting. This was motivated primarily by file size worries, cascade fits with many nodes and large numbers of samples (100+) can consume a lot of storage space.

Tests

Ran bin/run_xrst.sh and got no errors:

$ bin/run_xrst.sh
xrst --local_toc --target html --html_theme sphinx_rtd_theme --index_page_name at_cascade 
Using following input_files: git ls-files
sphinx-build -b html -j 1 build/rst build/html
rm -r build/html/_sources
cp -r build/rst/_sources build/html/_sources
xrst: OK
run_xrst.sh: OK
$

Notes

I expect this'll be another feature me & Nate try to tackle in the future, but wanted to add it to the list so it's tracked in the same place as all the other items. I'm also not 100% certain this will prove possible, eg if the predict samples depend on the fit samples, but need to know more about whether/how prediction samples depend on fit samples.

[bradbell]

It seems to me that you should just generate a small number of samples; e.g., 20 and then expand the predictions, when you use them by repeating each sample some multiple, e.g. 50, each time you use them.