Neptune Fetcher

Note

This package is experimental and only works with Neptune Scale (scale.neptune.ai), which is in beta.

For the Python API corresponding to app.neptune.ai, see neptune-client.

Neptune Fetcher is a read-only API for querying metadata logged with the Neptune Scale client. The separation makes it safer and more efficient to fetch data from Neptune.

With the Fetcher API, you can:

• List experiments, runs, and attributes of a project.
• Fetch experiment or run metadata as a data frame.
• Define filters to fetch experiments, runs, and attributes that meet certain criteria.

Documentation

• Fetching how-to guides
• Fetcher API reference
• Update your code from old Fetcher to Alpha

Installation

pip install -U neptune-fetcher

Set your Neptune API token and project name as environment variables:

export NEPTUNE_API_TOKEN="h0dHBzOi8aHR0cHM.4kl0jvYh3Kb8...ifQ=="

export NEPTUNE_PROJECT="workspace-name/project-name"

For help, see Get started in the Neptune documentation.

Note: To change the token or project, you can set the context directly in the code. This way, you can work with multiple projects at the same time.

Usage

Import the alpha module:

import neptune_fetcher.alpha as npt

To fetch experiment metadata from your project, use the fetch_experiments_table() function.

• To filter experiments to return, use the experiments parameter.
• To specify attributes to include as columns, use the attributes parameter.

npt.fetch_experiments_table(
    experiments=r"exp.*",
    attributes=r".*metric.*/val_.+",
)

           metrics/val_accuracy metrics/val_loss
                           last             last
experiment
exp_ergwq              0.278149         0.336344
exp_qgguv              0.160260         0.790268
exp_cnuwh              0.702490         0.366390
exp_kokxd              0.301545         0.917683
exp_gyvpp              0.999489         0.069839

To fetch values at each step, use fetch_metrics():

npt.fetch_metrics(
    experiments=r"exp.*",
    attributes=r".*metric.*/val_.+",
)

  experiment  step  metrics/val_accuracy  metrics/val_loss
0  exp_dczjz     0              0.432187          0.823375
1  exp_dczjz     1              0.649685          0.971732
2  exp_dczjz     2              0.760142          0.154741
3  exp_dczjz     3              0.719508          0.504652
4  exp_dczjz     4              0.180321          0.503800
5  exp_hgctc     0              0.012390          0.171790
6  exp_hgctc     1              0.392041          0.768675
7  exp_hgctc     2                  None          0.847072
8  exp_hgctc     3              0.479945          0.323537
9  exp_hgctc     4              0.102646          0.055511

You can define detailed criteria for which experiments to search or which attributes to return.

For instructions, see the how-to guides in the Neptune documentation:

• Listing project contents
• Fetching metadata
• Constructing filters
• Working with runs

---

Old Fetcher API

For documentation related to the previous version of the Fetcher API, see the docs/old/ directory:

• docs/old/usage.md
• docs/old/api_reference.md

To update your code to the new version, see Migrate to Fetcher Alpha in the Neptune documentation.

---

License

This project is licensed under the Apache License Version 2.0. For details, see Apache License Version 2.0.