Geo-distributed Model Deployment
Without building and configuring a complex Kubernetes Cluster. FedML®Deploy can deploy models to nodes located at multiple regions and manage the traffic routing automatically.
Heterogeneity Model Deployment
FedML®Deploy can deploy models to different types of devices, such as CPU, GPU, TPU, etc. Whether it is a single Macbook or a power A100, they can be connected together in an easy manner.
Autoscaling and Failover
FedML®Deploy can automatically scale up and down the number of replicas of the model deployment based on the QPS, CPU, GPU, and memory usage. It can also automatically recover from failures through backup nodes. Concat us for more details.