The Wayback Machine - https://web.archive.org/web/20170602175312/https://services.google.com/fb/forms/tpusignup/

Learn more about Cloud TPUs and the TensorFlow Research Cloud

Sign up below to get notified as more information about Cloud TPUs and the TensorFlow Research Cloud (TFRC) becomes available.

• First Name *
• Last Name *
• Email *
• Organization Name *
• Team Name
• Country *
• Which of the following interest you? *
  • Cloud TPUs
  • The TensorFlow Research Cloud (TFRC)
  • TPU Pods (large-scale ML supercomputers)
  • Other
• interest_other
• How many Cloud TPUs (180 teraflops each) would your team regularly use?
  • Don't know yet
  • 1 - 10
  • 10 - 50
  • 50 - 200
  • As many as we can get
  • Other
• regularuse_other
• What hardware do you currently use to train ML models?
  • P100 GPUs
  • Titan X GPUs
  • GTX 1080 GPUs
  • K80 GPUs
  • CPUs
  • Other special-purpose hardware
  • Decline to state
  • Other
• hardwaretraining_other
• What platforms do you use to train ML models?
  • None
  • A personal workstation
  • Small cluster of machines maintained directly by my team
  • Datacenters run by your organization
  • AWS
  • Azure
  • Google Cloud Platform
  • Softlayer
  • Nimbix
  • Other external cloud
  • Decline to state
  • Other
• workloadplatforms_other
• How long do your ML models typically take to train?
  • Minutes
  • Hours
  • Days
  • Weeks
  • Months
• How large are your typical training sets?
  • < 10 GB
  • 10 GB - 100 GB
  • 100 GB - 1 TB
  • 1 TB - 10 TB
  • 10 TB - 100 TB
  • 100 TB - 1 PB
  • > 1 PB
  • Other
• typicaltrainingsetsize_other
• Which categories of ML models are important to you?
  • Logistic Regression
  • Fully connected networks
  • Convolutional neural networks (CNNs)
  • Recurrent neural networks (RNNs)
  • Random forests
  • Reinforcement learning
  • Generative models
  • Other
• importantcategories_other
• At a high level, which of the following do you expect to do with Cloud TPUs?
  • Current ML tasks, but at a larger scale / higher speed
  • Current ML tasks, but at a lower cost
  • New ML tasks that are qualitatively different and / or infeasible now
  • Other
• highlevelexpect_other
• Communication Preferences *
  • Yes, please send me more information about the topics above
  • No thanks
• I accept Google's Terms and Conditions and acknowledge that my information will be used in accordance with Google's Privacy Policy.
• Sign up