Glossary

InfiniBand is a high-throughput, low-latency network architecture often used in data centers and supercomputing environments to link servers, storage, and GPUs.

DeepSpeed is a deep learning optimization library developed by Microsoft, designed to scale and speed up training of large models, particularly for transformer-based architectures.

Kubeflow is an open-source platform designed for managing machine learning (ML) workflows, particularly in Kubernetes environments. It allows for the easy deployment, monitoring, and scaling of ML models.