Running GPT-3 Models with NVIDIA® NeMo Megatron on Azure

This is a brief tutorial given by Hugo Affiticati, Technical Program Manager for the HPC and AI Benchmarking Team at Microsoft Azure, about how you can run GPT-3 model on Azure using NVIDIA NeMo Megatron.

This benchmark test tutorial uses the Azure NDm A100 v4 virtual machine, featuring eight (8) NVIDIA A100 80GB GPUs and 400GB/s InfiniBand networking interconnects. The Azure CycleCloud tutorial video (also available on YouTube) is also handy for context on this video.

For complete information and performance details, please go to https://aka.ms/AI/NeMo

#Microsoft #MicrosoftAzure