EADST

Quick Review: SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression

作者：XD / 发表： 2023年12月6日 23:57 / 更新： 2023年12月6日 23:57 / 科研学习 / 阅读量：168

SpQR: A Sparse-Quantized Representation for Near-Lossless Large Language Model Weight Compression

Core Approach:

  • GPTQ without Outliers: Focuses on eliminating outliers during the GPTQ process, enabling more efficient and accurate weight compression for large language models.

本文作者：XD 转载请标明出处：http://www.eadst.com/blog/225

本站采用知识共享署名-非商业性使用-相同方式共享 4.0 国际许可协议进行许可。

上一篇
Quick Review: Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs
下一篇
Quick Review: QUIK: Towards End-to-end 4-Bit Inference on Generative Large Language Models
相关标签
LLM Quantization
About Me
XD
Goals determine what you are going to be.
Category
标签云
VPN Attention XGBoost Pandas Bert XML torchinfo Base64 Michelin Template transformers uwsgi GPT4 Gemma GIT Website diffusers Shortcut Anaconda Git Zip Ubuntu llama.cpp Qwen Data Dataset Interview Use AI Streamlit Hungarian EXCEL SPIE OpenAI v2ray 腾讯云 阿里云 ChatGPT Excel Domain Color InvalidArgumentError C++ COCO Datetime OpenCV Plotly CAM Proxy ResNet-50 Jetson Heatmap PIP Python Ptyhon Land scipy Translation Review ModelScope Food TSV Docker 搞笑 Google Tensor Web 签证 Plate OCR BTC NLTK RGB Augmentation UNIX 域名 报税 HuggingFace LeetCode Github SQL LLM logger Quantization GPTQ Distillation Crawler Diagram Pickle Quantize Nginx PyCharm SQLite DeepStream WebCrawler CEIR Random Numpy CTC Safetensors ONNX LLAMA GoogLeNet BeautifulSoup LaTeX LoRA Django CV Input TensorFlow TensorRT Image2Text JSON Linux Pillow Knowledge PDF Algorithm Windows CUDA Sklearn Bitcoin VGG-16 Markdown Disk Baidu NLP Tiktoken Video QWEN Conda Cloudreve GGML Clash 关于博主 HaggingFace Statistics Paddle CSV Password SVR Pytorch uWSGI Bin FlashAttention Paper tar YOLO NameSilo Permission Tracking Bipartite
站点统计

本站现有博文242篇,共被浏览292073次

本站已经建立1782天!

热门文章
文章归档
回到顶部