This library contains tools that help you with:
  • logging/monitoring things in Neptune
  • visualizations of experiments and hyperparameters
  • running hyper parameter optimization in Neptune

Some cool functions are:


which lets you visualize learning curves on an interactive chart. You can choose which curves you want to view by clicking on the legend, select the x range on the top chart and view it in detail on the bottom.

from neptunecontrib.viz.experiment import channel_curve_compare


which lets you convert hyperopt.Trails object into scipy.optimize.OptimizeResults and visualize it with scikit-optimize diagnostic tools.

import neptunecontrib.hpo.utils as hp_utils
import neptunecontrib.monitoring.skopt as sk_monitor

results = hp_utils.hyperopt2skopt(trials, space)

sk_monitor.send_runs(results, ctx)
sk_monitor.send_best_parameters(results, ctx)
sk_monitor.send_plot_convergence(results, ctx)