paper-with-me

Papers

Fast NF4 Dequantization Kernels for Large Language Model Inference

2026-04-02 · Xiangbo Qi, Chaoyi Jiang, Murali Annavaram arxiv

Large language models (LLMs) have grown beyond the memory capacity of single GPU devices, necessitating quantization techniques for practical deployment. While NF4 (4-bit NormalFloat) quantization enables 4$\times$ memory reduction, inference on current NVIDIA GPUs (e.g., Ampere A100) requires expensive dequantization back to FP16 format, creating a critical performance bottleneck. This paper presents a lightweight shared memory optimization that addresses this gap through principled memory hierarchy exploitation while maintaining full ecosystem compatibility. We compare our technique against the open-source BitsAndBytes implementation, achieving 2.0--2.2$\times$ kernel speedup across three models (Gemma 27B, Qwen3 32B, and Llama3.3 70B) and up to 1.54$\times$ end-to-end improvement by leveraging the 12--15$\times$ latency advantage of shared memory over global memory access. Our optimization reduces instruction counts through simplified indexing logic while using only 64 bytes of shared memory per thread block, demonstrating that lightweight optimizations can deliver substantial performance gains with minimal engineering effort. This work provides a plug-and-play solution for the HuggingFace ecosystem that democratizes access to advanced models on existing GPU infrastructure.

📄 PDF Abstract BibTeX arXiv:2604.02556

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference

2024-02-15 · Taesu Kim, Jongho Lee, Daehyun Ahn, Sarang Kim 외

We introduce QUICK, a group of novel optimized CUDA kernels for the efficient inference of quantized Large Language Models (LLMs). QUICK addresses the shared memory bank-conflict problem of state-of-the-art mixed precisi…

GPUQuantization

LiquidGEMM: Hardware-Efficient W4A8 GEMM Kernel for High-Performance LLM Serving

2025-09-01 · Huanqi Hu, Bowen Xiao, Shixuan Sun, Jianian Yin 외 arxiv

Quantization is a critical technique for accelerating LLM inference by reducing memory footprint and improving computational efficiency. Among various schemes, 4-bit weight and 8-bit activation quantization (W4A8) offers…

Computational Efficiency

Fast Matrix Multiplications for Lookup Table-Quantized LLMs

2024-07-15 · Han Guo, William Brandon, Radostin Cholakov, Jonathan Ragan-Kelley 외

The deployment of large language models (LLMs) is often constrained by memory bandwidth, where the primary bottleneck is the cost of transferring model parameters from the GPU's global memory to its registers. When coupl…

Quantization

Fast and Efficient 2-bit LLM Inference on GPU: 2/4/16-bit in a Weight Matrix with Asynchronous Dequantization

2023-11-28 · Jinhao Li, Jiaming Xu, Shiyao Li, Shan Huang 외

Large language models (LLMs) have demonstrated impressive abilities in various domains while the inference cost is expensive. Many previous studies exploit quantization methods to reduce LLM inference cost by reducing la…

GPUQuantization

Multi-Scale Dequant: Eliminating Dequantization Bottleneck via Activation Decomposition for Efficient LLM Inference

2026-05-13 · Lingchao Zheng, Yuwei Fan, Jun Li, Chengqiu Hu 외 arxiv

Quantization is essential for efficient large language model (LLM) inference, yet the dequantization step-converting low-bit weights back to high-precision for matrix multiplication has become a critical bottleneck on mo…