Picture for Junda Feng

Junda Feng

ByteScale: Efficient Scaling of LLM Training with a 2048K Context Length on More Than 12,000 GPUs

Add code
Feb 28, 2025
Viaarxiv icon