How Databricks' FlashOptim cuts LLM training memory by 50 percent

Databricks researchers introduced FlashOptim, a suite of memory-optimization techniques for deep learning optimizers. FlashOptim slashes per-parameter memory consumption by more than 50 percent without sacrificing training throughput or model quality, enabling practitioners with limited hardware to train larger models.