Picture for Hyojung Lee

Hyojung Lee

FireQ: Fast INT4-FP8 Kernel and RoPE-aware Quantization for LLM Inference Acceleration

Add code
May 28, 2025
Viaarxiv icon