AI RESEARCH
ChunkFT: Byte-Streamed Optimization for Memory-Efficient Full Fine-Tuning
arXiv CS.LG
•
ArXi:2605.21177v1 Announce Type: new This work presents \textsc{ChunkFT}, a memory-efficient fine-tuning framework that reformulates full-parameter fine-tuning around a dynamically activated working set. \textsc{ChunkFT} enables gradient computation for arbitrary sub-tensors without modifying the network architecture, providing an algorithmic foundation for optimizing arbitrary sub-networks while avoiding standard dense gradient computation. We provide a theoretical convergence analysis of \textsc{ChunkFT} in the deterministic setting.