LLM Training on Run:ai with NVIDIA NeMo Megatron 💥

Omer Dayan shows how it feels to train large models on 32 A100 80GB GPUs with Run:ai. Optimized infra, optimized software stack, fastest training possible.
All on Kubernetes with Run:ai and NVIDIA’s NeMo Megatron
Thanks to our partners at NVIDIA for the collaboration on this one.