Hi PH! The starting point for Gradient was the revelation that people doing deep learning (DL) today were spending most of their thinking about infrastructure. This was a huge inefficiency that we found to be present across the board -- from researchers to DL teams at companies. To solve this problem, we set out to build an abstraction layer on top of raw infrastructure that allows developers to spend 100% of their time building models. In addition to the Notebooks, Jobs and CLI, we’re in the process of rolling out support for the TPU! This chip is a game changer for DL in terms of performance but is anything but trivial to get running (like most GPU infra, it doesn’t “just work”). The TPU really highlights the need for a software layer that bridges the gap between the power of accelerated chips and the people that just need this stuff to work. We have a ton of work to do, but we’re super excited to share our progress with you! We have a couple tutorials in our Knowledge Base (https://paperspace.zendesk.com/h... ) and also added a section to our Community (https://community.paperspace.com... ) as well. Let us know what you think and feel free to ping me with questions. ⚡

Paperspace is one of those companies that provides a glimpse into the future. They are powering game-changing advancements in tech across the board like self-driving cars, next-gen video games, and even moonshots like solving cancer with artificial intelligence.

<3

Love papersoace have been using it since beta days and it just keeps getting better. Tried the neural style transfer on it and it completed in 4 minutes. Awesome can’t wait to play more when I get out of work

Pros: Powerful job runner, Ability to manage multiple experiments Cons: You have to use PaperSpace's GPU Cloud, and can't run this on your own infra.

This is super cool!