LLM training code for Databricks foundation models

A Data Streaming Library for Efficient Neural Network Training

Making large AI models cheaper, faster and more accessible

An implementation of model parallel autoregressive transformers on GPUs, based on the Megatron and DeepSpeed libraries