S {2} FT: Efficient, Scalable and Generalizable LLM Fine-tuning by Structured Sparsity

Open in new window