https://junhyun.me/tag/llm-compression/LLM CompressionHugo Blox Builder (https://hugoblox.com)en-usSun, 01 Dec 2024 00:00:00 +0000https://junhyun.me/media/icon_hu97bcb6f0f3ad9a2f164a1f560b52571a_13652_512x512_fill_lanczos_center_3.pnghttps://junhyun.me/tag/llm-compression/https://junhyun.me/project/external-project/Sun, 01 Dec 2024 00:00:00 +0000https://junhyun.me/project/external-project/<p>Graduation project focused on improving AWQ (Activation-aware Weight Quantization) with extra scaling.</p> <ul> <li>Obtained lower perplexity for INT3-quantized OPT and Llama 2 models.</li> </ul>