Google Cloud launches TensorFlow Enterprise
Based on Google’s popular, open source TensorFlow machine learning library, TensorFlow Enterprise is positioned to help machine learning researchers accelerate the creation of machine learning and deep learning models and ensure the reliability of AI applications. Workloads in Google Cloud can be scaled and compatibility-tested.
Three features of TensorFlow Enterprise are highlighted:
- Enterprise-grade support, with long-term version support for TensorFlow. For certain versions, security patches and select bug fixes will be provided for as long as three years. These versions will be supported on Google Cloud, with patches and fixes accessible in the TensorFlow code repo. Also, “white-glove” service will be offered to cutting-edge customers, featuring engineer-to-engineer assistance from TensorFlow and Google Cloud teams at Google.
- Cloud-scale performance, with Google Cloud providing a range of compute options for training and deploying models. Deep Learning VMs, now generally available, and Deep Learning containers, in beta, are featured. Both products are available for Nvidia GPUs and Google’s Cloud TPU, for AI. TensorFlow Enterprise optimizations, meanwhile, have increased data reading times by as much as three times.
- Managed services, with enterprises able to leverage cloud services such as Kubernetes Engine and AI Platform.
Users can get benefits of TensorFlow Enterprise by using the TensorFlow Enterprise Distribution on AI Platform Notebooks, AI Platform Deep Learning Containers, and the AI Platform Deep Learning VM Image.
How to access TensorFlow Enterprise
A free trial of TensorFlow Enterprise is available on the Google Cloud website. The service is currently in a beta stage.