Use charts instead of console logsInstead of using console logs, figure out what’s going on in your training by looking at the charts.
Use Jupyter Notebooks in the cloudPrototype interactively using Jupyter Notebooks in the cloud. Neptune saves your code and outputs automatically.
Quickly find and compare your best experimentsUse your private “Kaggle leaderboard” like a dashboard to filter, explore and sort through your experiments. Find your top machine learning models based on your favorite metric.
Optimize hyperparameters automaticallyInstead of manually tuning the hyperparameters of your machine learning models or to avoid tuning altogether, utilize cloud resources and automatically find the optimal parameters.
Neptune comes with a grid search optimization engine that can tune your code’s parameters regardless of the machine learning library you’re using.
$ neptune send train_cnn.py
Take advantage of leading-edge hardwareWant to use NVIDIA® Tesla® K80 GPUs to train your deep learning models? Or Google’s infrastructure for your computations? It’s a snap! Just select your machine type and send your computations to the cloud.
$ neptune send --worker gcp-medium
$ neptune send --worker gcp-gpu-large
Pay per secondNeptune comes with a data science-friendly, per-second billing model. Forget about expensive hourly rates and pay only for what you use - the exact time of your experiment run.
Use pre-configured compute environmentsTired of manually configuring your remote machines or installing missing libraries?
Don’t waste your time on DevOps work! Use one of Neptune’s pre-configured environments (Docker images) and focus on data science instead.
$ neptune send --environment tensorflow
$ neptune send --environment keras
$ neptune send --environment theano