The Workflow Features Try us
Build better neural nets. Faster.
Keplr Helps You
Identify the right deep learning model for your dataset
Parameter Space
Fast and powerful
Model selection workflow in the cloud
Launch an experiment
From your command line, from a config file or any keras model
We launch a dedicated cluster
Train and score models in your parameter space
Watch your models train
See how different configurations in parameter space perform on your dataset in realtime
Download the models you like
And get a full report at the end of your experiment.
Keplr also helps you
optimize existing models
cusp generate
Keplr generates a hyperparameter space around your model
cusp run my_model
Arrive at a better model in this space
Built for data scientists and developers,
like ourselves.
Optimize anything in a neural network
Automatically explore model architectures, layer parameters and training parameters.
Fast, configurable, auto-scaling hardware
Only pay for as much memory/storage/GPU as you need. Monitor hardware metrics in realtime.
Intelligent, Transparent Hyperparameter Optimization
Track and control the progress of the optimization process, stop any test at any time.
Deep Learning Visualizations
Get insights and intuition about what makes good models good.
Supercharge your deep learning workflow