Welcome to the rubicon_ml examples!¶
Try out rubicon_ml for yourself without any setup required! If you'd like to learn more about
rubicon_ml before getting started, check out the
documentation on GitHub.
Exploring an existing rubicon_ml project¶
Below we'll explore an existing rubicon_ml project using the built-in visualization module to
inspect and analyze some previously trained models.
This GitHub repository contains
a simple classification model in my_model/my_model.py. Over three commits, three
different types of models were used to classify species of penguins by
some physical characteristics. During each model run, rubicon_ml was used to log the results.
The rubicon_ml logs from that project have been copied locally into this Binder session. In
a more realistic use case, you may be reading shared rubicon_ml logs from S3.
There are a number of questions you'll be able to answer by using the experiment table below:
- What were the input parameters and output metrics of each model run?
- What type of model did each run use?
- Where is the source code that was used to generate each model run's logs?
- Which model and input parameters produced the best results?
Note: default_server_url should match the root of the URL currently in your address
bar - change as necessary.
import os
from rubicon_ml import Rubicon
from rubicon_ml.viz import ExperimentsTable
default_server_url = "https://hub.gke2.mybinder.org/"
# default_server_url = "https://hub-binder.mybinder.ovh"
port = 8050
requests_pathname_prefix = f"{os.environ['JUPYTERHUB_SERVICE_PREFIX']}proxy/{port}/"
dash_kwargs = {"requests_pathname_prefix": requests_pathname_prefix}
run_server_kwargs = {"proxy": f"http://127.0.0.1:{port}::{default_server_url}", "port": port}
rubicon = Rubicon(persistence="filesystem", root_dir=f"{os.getcwd()}/rubicon-root")
project = rubicon.get_project(name="Classifying Penguins")
experiments_table = ExperimentsTable(experiments=project.experiments())
experiments_table.show(dash_kwargs=dash_kwargs, run_server_kwargs=run_server_kwargs)
Running Other rubicon_ml Examples¶
You can run any of the examples from the rubicon_ml docs in this Binder session! These examples
will show how to log your own projects and experiments, share and visualize them. Run the cell below
to generate the links to these examples for the current Binder session.
Note: The examples in the "Visualizations" section will need to be manually updated to provide
the show and serve methods the same dash_kwargs and run_server_kwargs arguments as the cell
above.
session_base_url = os.path.dirname(os.path.dirname(os.path.dirname((
default_server_url + requests_pathname_prefix
))))
print(f"Quick Look: \t{session_base_url}/tree/notebooks/quick-look")
print(f"Examples: \t{session_base_url}/tree/notebooks/logging-examples")
print(f"Integrations: \t{session_base_url}/tree/notebooks/integrations")
print(f"Visualizations: {session_base_url}/tree/notebooks/viz")