Distributed LLM Pretraining Torchtitan

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

math-inc Updated

File contents

math-inc/opengauss commit b1777ab3b9

Frequently asked questions

npx get-skill add math-inc/distributed-llm-pretraining-torchtitan