DomWhisper

排行榜
登录注册
arxiv.org favicon

[2305.14314] QLoRA: Efficient Finetuning of Quantized LLMs

https://arxiv.org/abs/2305.14314
7
公开标注数
6
参与人数
2026-07-22 10:05:02
首次 Whisper

本页的公开 Whisper

arxiv.org favicon
arxiv.org
划选高亮2026-08-11 18:38:31
原文高亮摘录
“QLoRA backprop”
Whisper 随想笔记
My 48GB card finally feels useful for something other than gaming.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 16:29:02
原文高亮摘录
“arXivLabs is a framework that allows collaborators to develop and share new arXiv features”
Whisper 随想笔记
Sounds like a cool way for the community to pitch in on features.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 13:26:02
原文高亮摘录
“preserving full 16-bit finetuning task performance”
Whisper 随想笔记
I just hope it's not another paper that overpromises on benchmarks.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 13:17:02
原文高亮摘录
“preserving full 16-bit finetuning task performance”
Whisper 随想笔记
Sounds too good to be true, but if it works, that's a game changer.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 10:23:02
原文高亮摘录
“reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU”
Whisper 随想笔记
Tried it on a 13B model, memory drop was insane. Game changer for my lab.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 10:14:02
原文高亮摘录
“reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU”
Whisper 随想笔记
But does it really keep full 16-bit performance? I'd like to see more tests.
arxiv.org favicon
arxiv.org
划选高亮2026-07-22 10:05:02
原文高亮摘录
“reduces memory usage enough to finetune a 65B parameter model on a single 48GB GPU”
Whisper 随想笔记
Finally can train big models without selling a kidney for GPUs.

分享本页 Whisper

分享到 X
短链接
https://domwhisper.com/s/2e95ee3ff8d6
嵌入代码
<iframe src="https://domwhisper.com/embed/2e95ee3ff8d6" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>

看看大家在 arxiv.org 上讨论了什么

安装 DomWhisper，浏览网页时实时查看 whisper，也可以加入讨论。

获取插件