TGViewer
肾小管 肾小管 @shengxiaoguan · 56 subscribers
Post #683 633
https://arxiv.org/abs/2402.17764
1.58bit (ternary parameters 1,0,-1) LLMs, showing performance and perplexity equivalent to full fp16 models of same parameter size. Implications are staggering. Current methods of quantization obsolete. 120B models fitting into 24GB VRAM. Democratization of powerful models to all with consumer GPUs.
https://www.reddit.com/r/LocalLLaMA/comments/1b21bbx/this_is_pretty_revolutionary_for_the_local_llm/
arXiv.org The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits Recent research, such as BitNet, is paving the way for a new era of 1-bit Large Language Models (LLMs). In this work, we introduce a 1-bit LLM variant, namely BitNet b1.58, in which every single...
  • ❤ 2
  • 👍 1
Open in Telegram All posts from @shengxiaoguan
More from @shengxiaoguan
  1. Aug 3, 2025（转自qq空间，已查证 土豆它妈是番茄…… 手里的薯条番茄酱突然变成亲子丼了……… 中国科学家证实番茄是土豆的杂交母本，成果于7月31日发表于细胞Cell 原文：Ancient h…
  2. Jul 9, 2024https://lllyasviel.github.io/pages/paints_undo/ AI也会一步步画画喽 虽然类似于一步步撤销
  3. Jun 16, 2023https://t.me/loveuchangeless/4210 https://t.me/loveuchangeless/4211 https://t.me/loveuchan…
  4. Jun 14, 2023https://omnimotion.github.io/ Tracking Everything Everywhere All at Once 实时跟随！！！
  5. May 19, 2023video post
  6. May 19, 2023https://github.com/XingangPan/DragGAN https://vcai.mpi-inf.mpg.de/projects/DragGAN/ GAN杀回来…
Threads Profile ViewerView any public Threads profile without an account.Open ThreadLook →Writing with AI? Make it sound human.Metric37 rewrites AI drafts so they read naturally. Free AI detector, 1,500 words free.Try Metric37 →