SwiftKV: Fast Prefill-Optimized Inference with Knowledge-Preserving Model Transformation Paper • 2410.03960 • Published Oct 4, 2024 • 1 • 2