Update installation docs for Prophet

[ParthivNaresh]

Fixes #2688

[codecov]

Codecov Report

Merging #2713 (27d5e21) into main (6ae6c16) will increase coverage by 0.8%.
The diff coverage is n/a.

@@           Coverage Diff           @@
##            main   #2713     +/-   ##
=======================================
+ Coverage   99.2%   99.9%   +0.8%     
=======================================
  Files        301     301             
  Lines      27827   27827             
=======================================
+ Hits       27581   27778    +197     
+ Misses       246      49    -197     

Impacted Files	Coverage Δ
evalml/tests/automl_tests/test_automl.py	99.7% <0.0%> (+0.1%)	⬆️
evalml/automl/utils.py	100.0% <0.0%> (+1.7%)	⬆️
...lml/tests/automl_tests/test_iterative_algorithm.py	100.0% <0.0%> (+4.5%)	⬆️
evalml/tests/automl_tests/test_automl_utils.py	100.0% <0.0%> (+9.4%)	⬆️
.../automl_tests/test_automl_search_classification.py	100.0% <0.0%> (+10.6%)	⬆️
...ests/automl_tests/test_automl_search_regression.py	100.0% <0.0%> (+15.5%)	⬆️
evalml/automl/pipeline_search_plots.py	100.0% <0.0%> (+53.6%)	⬆️
...l/tests/automl_tests/test_pipeline_search_plots.py	100.0% <0.0%> (+74.5%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6ae6c16...27d5e21. Read the comment docs.

[gsheni]

You might want to think about moving this instructions to Install page on the docs. This currently will make the README.md much longer. You could link from the README.md to the Install documentation page.

We put graphviz install instructions in our Install docs:
https://featuretools.alteryx.com/en/latest/install.html#installing-graphviz

[ParthivNaresh]

@gsheni Thanks for the recommendation!

[freddyaboulton]

I also agree with this recommendation @ParthivNaresh !

[freddyaboulton]

When I first tried this out in a fresh environment, I got the infamous no such file ...prophet_model.bin error when I tried running the unit tests.

I was really confused about why this happened, but when I looked at the pip install, I saw:

Collecting prophet==1.0.1
  Using cached prophet-1.0.1-py3-none-any.whl

This stood out to me because as we all know by now, prophet does not publish any wheels! So I was confused why we weren't building from source.

Then I read the pip documentation, and I learned that locally built wheels are added to the cache.

So my guess as to what went wrong for me is that I had installed prophet beforehand without a cmdstan backend and then subsequent installs with a cmdstan backend were using the cached version so it was not properly setting up cmdstan.

So in short, I think we need to do pip install --no-cache-dir prophet==1.0.1 instead. That properly set up the cmdstan backend and I was able to get the unit tests to pass.

Apparently, --no-cache-dir is discouraged, but the workaround in the pip docs did not work for me.

[angela97lin]

As a beginner user, I'm a bit confused on how to use this command. This assumes that a user actually has the evalml repo; that is, I can't just create a new virtual env, pip install evalml, and then run this. I'd need to have cloned the evalml repo and run this while in the evalml repo, correct?

If that's true, we should make that clear! (though also makes me question how often people would take this route)

[chukarsten]

This seemed clear and it worked for me.

[freddyaboulton]

Thank you @ParthivNaresh !