Skip to content

Sacred-compatible UI for experiment tracking. ๐Ÿ“Š Log and visualize machine learning model metrics, hyperparameters, code, dataset versions, and more.

License

Notifications You must be signed in to change notification settings

neptune-ai/neptune-sacred

Folders and files

NameName
Last commit message
Last commit date

Latest commit

ย 

History

85 Commits
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 
ย 

Repository files navigation

Neptune + Sacred Integration

Neptune is a lightweight experiment tracker that offers a single place to track, compare, store, and collaborate on experiments and models.

This integration lets you use Neptune as a UI for the experiments you track in Sacred.

What will you get with this integration?

  • Log, organize, visualize, and compare ML experiments in a single place
  • Monitor model training live
  • Version and query production-ready models and associated metadata (e.g., datasets)
  • Collaborate with the team and across the organization

What will be logged to Neptune?

  • Hyperparameters
  • Losses and metrics
  • Training code (Python scripts or Jupyter notebooks) and Git information
  • Dataset version
  • Model configuration
  • Other metadata

image

Resources

Example

On the command line:

pip install neptune-sacred

In Python:

import neptune

# Start a run
run = neptune.init_run(
    project = "common/sacred-integration",
    api_token = neptune.ANONYMOUS_API_TOKEN,
)

# Create a Sacred experiment
experiment = Experiment("image_classification", interactive=True)

# Add NeptuneObserver and run the experiment
experiment.observers.append(NeptuneObserver(run=run))
experiment.run()

Support

If you got stuck or simply want to talk to us, here are your options:

  • Check our FAQ page
  • You can submit bug reports, feature requests, or contributions directly to the repository.
  • Chat! When in the Neptune application click on the blue message icon in the bottom-right corner and send a message. A real person will talk to you ASAP (typically very ASAP),
  • You can just shoot us an email at [email protected]