Optuna integration API reference

API reference for the callback and utilities of the Neptune-Optuna integration.

The Neptune-Optuna integration provides aNeptuneCallbackclass and two utility functions.

NeptuneCallback

A callback that logs metadata from an Optuna study to Neptune.

With this callback, you can log and display:

• Values and parameters for each trial.
• Current best values and parameters for the study.
• Visualizations from the optuna.visualizations module.
• Parameter distributions for each trial.
• The study object itself, to load it later.

Parameters

• If you pass an integer k , plots will be updated every k iterations.

• If you pass the string "never" , plots will not be logged.

• If you pass an integer k , the study will be updated every k iterations.

• If you pass the string "never" , the study will not be logged.

If your study is not multi-objective, this plot is not logged.

If your study is not using pruners, this plot is not logged.

Examples

Create a Neptune run and initialize a Neptune callback:

import neptune
import neptune.integrations.optuna as npt_utils

run = neptune.init_run()
neptune_callback = npt_utils.NeptuneCallback(run)

As a best practice, you should save your Neptune API token and project name as environment variables:

export NEPTUNE_API_TOKEN="h0dHBzOi8aHR0cHM6Lkc78ghs74kl0jv...Yh3Kb8"

export NEPTUNE_PROJECT="ml-team/classification"

Alternatively, you can pass the information when using a function that takesapi_tokenandprojectas arguments:

run = neptune.init_run(
 api_token="h0dHBzOi8aHR0cHM6Lkc78ghs74kl0jv...Yh3Kb8", # (1)!
 project="ml-team/classification", # (2)!
)

1. In the bottom-left corner, expand the user menu and select Get my API token .
2. You can copy the path from the project details ( → Details & privacy ).

If you haven't registered, you can log anonymously to a public project:

api_token=neptune.ANONYMOUS_API_TOKEN
project="common/quickstarts"

Make sure not to publish sensitive data through your code!

Or optionally pass a list of objective names:

neptune_callback = npt_utils.NeptuneCallback(
 run,
 target_names=["accuracy"],
)

neptune_callback = npt_utils.NeptuneCallback(
 run,
 target_names=["FLOPS", "accuracy"],
)

Log single and multi objective study metadata to Neptune by passing the callback to the Optuna study:

study = optuna.create_study(direction="maximize")
study.optimize(objective, n_trials=5, callbacks=[neptune_callback])

log_study_metadata()

Logs the metadata from an Optuna study to Neptune.

With this function, you can log and display:

• Values and parameters for each trial.
• Current best values and parameters for the study.
• Visualizations from the optuna.visualizations module.
• Parameter distributions for each trial.
• The study object itself, to load it later.

Parameters

The objects that are logged depend on the study storage type used: If InMemoryStorage is used, the pickled study object will be logged to Neptune. Otherwise the database URL will be logged.

If your study is not multi-objective, this plot is not logged.

If your study is not using pruners, this plot is not logged.

Examples

Create a Neptune run:

import neptune
import neptune.integrations.optuna as npt_utils

run = neptune.init_run(project="workspace-name/project-name") # (1)!

1. The full project name. For example,"ml-team/classification".
2. You can copy the name from the project details ( → Details & privacy )
3. You can also find a pre-filled project string in Experiments → Create a new run .

The full project name. For example,"ml-team/classification".

Create and run the study:

study = optuna.create_study(direction="maximize")
study.optimize(objective, n_trials=5)

Log single and multi objective study metadata to Neptune:

import neptune.integrations.optuna as npt_utils

npt_utils.log_study_metadata(study, run)

Or optionally pass a list of objective names:

npt_utils.log_study_metadata(
 study,
 run,
 target_names=["accuracy"],
)

npt_utils.log_study_metadata(
 study,
 run,
 target_names=["FLOPS", "accuracy"],
)

load_study_from_run()

Loads Optuna study from an existing Neptune run.

Loading mechanics depend on the study storage type used during the run:

• If the study used InMemoryStorage, it will be loaded from the logged pickled study object.
• If the study used database storage, it will be loaded from the logged database URL.

Parameters

Returns

An Optuna study object.

Examples

Initialize an existing run by passing the run ID:

import neptune

run = neptune.init_run(run="PRO-123")

The Neptune ID is a unique identifier for the run. TheExperimentstab displays it in the leftmost column.

In the run structure, the ID is stored in the system namespace (sys).

• If the run is active, you can obtain its ID withrun["sys/id"].fetch(). For example: >>>run=neptune.init_run()...>>>run["sys/id"].fetch()'CLS-26'
• If you set a custom run ID, it's stored atsys/custom_run_id: >>>run["sys/custom_run_id"].fetch()'vigilant-puffin-20bt9'

If the run is active, you can obtain its ID withrun["sys/id"].fetch(). For example:

>>> run = neptune.init_run()
...
>>> run["sys/id"].fetch()
'CLS-26'

If you set a custom run ID, it's stored atsys/custom_run_id:

>>> run["sys/custom_run_id"].fetch()
'vigilant-puffin-20bt9'

Load the study from the run and continue optimization:

import neptune.integrations.optuna as npt_utils

study = npt_utils.load_study_from_run(run)
study.optimize(objective, n_trials=20)

See also

neptune-optuna repo onGitHub

Related Documentation

• Neptune Package
• Run API

This page is originally sourced from the legacy docs.