SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models

Authors: Guangxuan Xiao, Ji Lin, Mickael Seznec, Hao Wu, Julien Demouth, Song Han (2022)

arXiv: 2211.10438

Domains

Inference

TLDR (English)

Moves activation outliers to weights through equivalent mathematical transformation, making INT8 inference feasible. Key engineering discovery enabling GPU FP8/INT8 deployment.

TLDR（中文）

把激活的 outlier 通过等价数学变换"挪"到权重上，使得 INT8 推理可行。是 GPU FP8/INT8 部署能 work 的关键工程发现。

Appears in These Articles

• KV Cache 与量化：让大模型跑得更快
• KV Cache and Quantization: Making Large Models Faster

Co-cited Papers

These papers appear in the same articles as this one

• LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale Tim Dettmers et al. (2022)
• QLoRA: Efficient Finetuning of Quantized LLMs Tim Dettmers et al. (2023)
• GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers Elias Frantar et al. (2022)
• Efficient Memory Management for Large Language Model Serving with PagedAttention Woosuk Kwon et al. (2023)

Related Papers

Other papers in the same domain

• DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model DeepSeek-AI (2024)
  ArchitectureInference
• FlashAttention-3: Fast and Accurate Attention with Asympotic IO Complexity Jay Shah et al. (2024)
  Inference
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters Charlie Snell et al. (2024)
  ReasoningInference
• GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints Joshua Ainslie et al. (2023)
  InferenceArchitecture
• Accelerating Large Language Model Decoding with Speculative Sampling Charlie Chen et al. (2023)
  Inference
• FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning Tri Dao (2023)
  Inference