Accelerating open machine learning research with Cloud TPUs

  • Our goal is to ensure that the most promising researchers in the world have access to enough compute power to imagine, implement, and publish the next wave of ML breakthroughs.
  • We’re setting up a program to accept applications for access to the TensorFlow Research Cloud and will evaluate applications on a rolling basis.
  • The program will be highly selective since demand for ML compute is overwhelming, but we specifically encourage individuals with a wide range of backgrounds, affiliations, and interests to apply.
  • The program will start small and scale up.

Researchers need enormous computational resources to train the machine learning models that have delivered
recent advances in medical imaging, speech recognition, game playing, and many other domains. The TensorFlow
Research Cloud is a cluster of 1,000 Cloud TPUs that provides the machine learning research community with
a total of 180 petaflops of raw compute power — at no charge — to support the next wave of breakthroughs.

@fchollet: If you are a researcher, you can apply now to start accessing Google Cloud TPUs for training your models –for free!

Researchers need enormous computational resources to train the machine learning models that have delivered recent advances in medical imaging, speech recognition, game playing, and many other domains. The TensorFlow Research Cloud is a cluster of 1,000 Cloud TPUs that provides the machine learning research community with a total of 180 petaflops of raw compute power — at no charge — to support the next wave of breakthroughs.

Google’s first-generation TPUs made it dramatically faster to run ML models that had already been trained, but the training had to be performed separately. Google’s second-generation Cloud TPUs are even more powerful, designed from the ground up to accelerate training ML models as well as running them. The TensorFlow Research Cloud provides researchers with access to these second-generation Cloud TPUs, each of which provides 180 teraflops of ML acceleration.

Our goal is to ensure that the most promising researchers in the world have access to enough compute power to imagine, implement, and publish the next wave of ML breakthroughs. We’re setting up a program to accept applications for access to the TensorFlow Research Cloud and will evaluate applications on a rolling basis. The program will be highly selective since demand for ML compute is overwhelming, but we specifically encourage individuals with a wide range of backgrounds, affiliations, and interests to apply. The program will start small and scale up.

Participants in the TFRC program will be expected to share their TFRC-supported research with the world through peer-reviewed publications, open-source code, blog posts, or other means. They should also be willing to share detailed feedback with Google to help us improve the TFRC program and the underlying Cloud TPU platform over time.

Machine learning researchers around the world have done amazing things with the limited computational resources they currently have available. We’d like to empower researchers to think even bigger and tackle exciting new challenges that would be inaccessible otherwise. We hope the TensorFlow Research Cloud will allow people from many different backgrounds to expand the frontier of machine learning research with new discoveries!

Accelerating open machine learning research with Cloud TPUs