Features

Distributed computing

 

You probably know that training deep learning models is faster — often order(s) of magnitude faster — when parallelized and distributed across many GPU workers. TensorPort’s infrastructure is capable of running your experiments at huge scale, on terabytes of data with hundreds of GPU workers. Our streamlined project creation process makes it easy to quickly set up multiple distributed experiments to run in parallel.

With TensorPort, you can be sure your code is running on the best available hardware at the best possible price. We bill compute usage by the minute at prices lower than any major cloud computing provider (feel free to compare our prices yourself!) and offer built-in timing tools making it easy for you to control your spending. Take a look at all our charges here if you’d like.

Reproducibility

 

With TensorPort, you can stop worrying about misplacing your data or losing track of changes to your code. TensorPort uses Git and Git Large File Storage (LFS) to version all of your files while maintaining maximum performance. We make it easy to train and test any version of an uploaded model using any version of uploaded datasets, so that you can always go back and reproduce your earlier experiments.

Collaboration

 

TensorPort gives your team the ability to easily collaborate on machine learning projects — you won't be able to imagine living without it! TensorPort is built for machine learning teams: we make it simple to invite others, share projects, monitor project activity, and manage permissions. The platform is fully integrated with Git so you can work with teammates on code and data the way you’re used to.

Flexibility

 

You are in charge! Our platform is not only easy to use, but also flexible and fully scalable no matter what the size of your projects. We believe that you understand your data best, so you have full control on how to build and train your models. TensorPort is NOT Yet Another Black Box Artificial Intelligence API (YABBAIAPI).

We have multiple subscription levels so you can choose the level of support and team size that best fits your needs, and we offer on-premise and private cloud TensorPort implementations as well as TensorPort cloud accounts.

Effortless! TensorPort is designed by AI researchers for AI researchers. We built it so everything works great without additional effort.

Integration

 

We like TensorFlow! TensorFlow is a great computational framework based on design, performance, flexibility and portability. TensorPort works flawlessly with TensorFlow. We also integrate TensorBoard, hyperparameter tuning, notifications, and other workflow tools to eliminate hurdles that will slow down your development.

You can use TensorPort through our graphical interface, your command line, our Python client or our open API, allowing you to utilize the platform in a way that fits with your existing workflow and productivity solutions.

Model Serving

 

Coming soon: We’re working on model deployment tools so that you’ll not only be able to run experiments with TensorPort, but also serve your trained models in the cloud in just a few clicks.

TensorPort is a product of Good AI Lab. We are a global team of scientists and engineers who aim to bring the best tools and practices to machine learning teams.