API reference

•  neptune

  ---

  Initialize new or existing Neptune objects.

  • init_run()
  • init_model()
  • init_model_version()
  • init_project()

•  management

  ---

  Manage your workspace and users.

  • invite_to_workspace()
  • create_project()
  • trash_objects()
  • more →

•  Field types

  ---

  A field is a location where metadata is logged. The field type determines the available methods and display options.

  • Artifact
  • File
  • FileSeries
  • FloatSeries
  • more →

•  Environment variables

  ---

  Use environment variables to make your setup more smooth and secure.

  • NEPTUNE_API_TOKEN
  • NEPTUNE_PROJECT
  • NEPTUNE_CUSTOM_RUN_ID
  • NEPTUNE_MODE
  • NEPTUNE_DATA_DIRECTORY
  • more →

•  Connection modes

  ---

  Set the logging and synchronization mode.

  • async (default)
  • debug
  • offline
  • read-only
  • sync

•  CLI

  ---

  You can use the Neptune Command Line Interface to manage and sync locally stored metadata.

  • neptune clear
  • neptune status
  • neptune sync

Browse the complete list of functions, parameters, and constants in the index →

Neptune client library

The Neptune Python API has the following structure:

• Packages:
  • neptune: Initialization functions for each type of Neptune object.
  • management: Functions for managing users and project contents.
• Modules:
  • utils: Utility functions for advanced use.
• Classes (Neptune objects):
  • Run: Designed to log experimentation metadata. Automatically tracks some metadata once initialized.
  • Model: Enables tracking of model metadata separately from experimentation runs.
  • ModelVersion: Represents a specific version of a model and has a lifecycle stage.

  • Project: Exposes fetching methods for run and model tables.

    Also useful for storing metadata that applies to the entire project, as a single source of truth.

View client library on GitHub 

Integrations

API references for Neptune integrations with other ML frameworks:

• Apache Airflow
• Catalyst
• Detectron2
• fastai
• Transformers
• Kedro
• Keras
• LightGBM
• Lightning
• MLflow
• Optuna
• Prophet
• PyTorch
• PyTorch Ignite
• Sacred
• scikit-learn
• skorch
• TensorBoard
• XGBoost

Errors, warnings, and other messages

Not foundInvalid, malformed, or unsupportedLimit exceededOther

• ModuleNotFoundError: No module named 'neptune'
• NVML Error: NVML Shared Library Not Found

• Out of range float values are not JSON compliant → Downgrade simplejson to 3.18
• X-coordinates (step) must be strictly increasing for series attribute
• Value of float attribute cannot be malformed
• 'Handler' object has no attribute 'append'
• TypeDoesNotSupportAttributeException
• NeptuneUnsupportedType

• NeptuneLimitExceedException
• NeptuneFieldCountLimitExceedException
• Max string attribute length exceeded

• WritingToArchivedProjectException
• Still waiting for the remaining X operations
• HTTPTooManyRequests

See also

• NQL (Neptune Query Language) – used to construct complex queries when fetching runs.
• System namespace (sys) – contains the name, description, tags, creation time, and other auto-logged metadata.
• Change the location of the .neptune folder – see how to move the directory or access it from elsewhere in your system.
• Set up a Neptune error handling function
• Help ≫ Debugging options