simplify and fix optuna-hpo notebook #404
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
| *.csv | ||
| *.csv.gz | ||
| *.csv.zip | ||
| *.tar.gz | ||
| *.zip |
There was a problem hiding this comment.
The rapids-optuna-hpo example ends up downloading a .zip file, which contains some inner zipped CSV files.
That's why I added .zip and .csv.zip here, and then though "might as well as some other common archives too", as I don't think we ever want any of these in source control.
| "metadata": {}, | ||
| "outputs": [], | ||
| "source": [ | ||
| "## Run this cell to install optuna\n", | ||
| "# !pip install optuna" | ||
| "#!pip install optuna optuna-integration" |
There was a problem hiding this comment.
This notebook calls the following Python code:
storage = optuna.integration.DaskStorage()The optuna.integration module is provided by a separate package, optuna-integration, which is not pulled in by default by pip install optuna.
Adding this avoids a ModuleNotFoundError later on in the notebook.
| "Best params: {'C': 1.486491072441749, 'penalty': 'l2', 'fit_intercept': True}\n", | ||
| "Number of finished trials: 144\n", | ||
| "```" | ||
| ] |
There was a problem hiding this comment.
The logs generated above:
- are huge
- contain stacktraces with absolute paths from wherever the code was run
Proposing here to instead clear the output of that cell and just keep this shorter snippet of the logs in a markdown block.
Contributes to #402 .
I ran through
source/examples/rapids-optuna-hpo/notebook.ipynbtoday. This proposes a few changes to that notebook based on that experience.Added bonus... I tested this with Python 3.11 (for #377).
Notes for Reviewers
I've left comments inline explaining the changes.