-
-
Categories
-
-
-
For Vendors
-
- For Experts
- Press Release
- Sign in / Sign up
-
Google Cloud ML Engine provide training and prediction services, which can be used together or individually. It has been used by enterprises to solve problems ranging from identifying clouds in satellite images, ensuring food safety, and responding four times faster to customer emails. The training and prediction services within ML Engine are now referred to as AI Platform Training and AI Platform Prediction.
Key Features:
Custom container support
Google can run any other framework on Cloud ML Engine along with native support for popular frameworks like TensorFlow. It simply upload a Docker container with training program and Cloud ML Engine that put it to work on Google's infrastructure.
Distributed training
Cloud ML Engine automatically sets up an environment for XGBoost and TensorFlow to run on multiple machines. It gets the speed that is needed by adding multiple GPUs to training job or splitting it across multiple VMs.
Automatic resource provisioning
Cloud ML Engine is a managed service which automates all resource provisioning and monitoring builds models using managed distributed training infrastructure that supports CPUs, GPUs, and TPUs; and accelerates model development by training across many nodes or running multiple experiments in parallel.
HyperTune
It achieves quick results by automatically tuning deep learning hyperparameters with HyperTune. HyperTune saves many hours of tedious and error-prone work.
Portable models
The open source TensorFlow SDK or other supported ML frameworks train models locally on sample datasets and use the Google Cloud Platform for training at scale. Models trained using Cloud ML Engine can be downloaded for local execution or mobile integration. It can also import scikit-learn, XGBoost, Keras, and TensorFlow models that have been trained anywhere for fully-managed, real-time prediction hosting.
Server-side pre-processing
It push deployment pre-processing to Google Cloud with scikit-learn pipelines and tf.transform. It can send raw data to models in production and reduce local computation, while also preventing data skew from being introduced through different pre-processing in training and prediction.
Integrated
Cloud ML Engine has deep integration with managed notebook service and data services for machine learning. Cloud Dataflow for feature processing, BigQuery for dashboard support and analysis, and Cloud Storage for data storage.
Multiple Frameworks
Training and Online Prediction support multiple frameworks to train and serve classification, regression, clustering, and dimensionality reduction models.
Analyst
Buyer, Healthcare, SME
James Smith