Category: AWS Inferentia

When choosing the infrastructure for their ML workloads, startups should consider how to best approach training and inference. Training is process by which a model is built and tuned for a specific task by learning from existing data. Inference is the process of using that model to make predictions based on new input data. Over the last five years, AWS has been investing in our own purpose-built accelerators to push the envelope on performance and compute cost for ML workloads. AWS Trainium and AWS Inferentia accelerators enable the lowest cost for training models and running inference in the cloud.