Keep on to blur preview images; turn off to show them clearly
CLM: Removing the GPU Memory Barrier for 3D Gaussian Splatting Abstract (excerpt): In this paper, we describe CLM, a system that allows 3DGS to render large scenes using a single consumer-grade GPU, e.g., RTX4090. It does so by offloading Gaussians to CPU memory and loading them into GPU memory only when necessary. To improve performance and reduce communication overheads, CLM uses a novel offloading strategy based on insights into 3DGS’s memory access patterns. This strategy enables efficient pipelining, which overlaps GPU-to-CPU communication, GPU computation, and CPU computation. Furthermore, CLM exploits these access patterns to reduce communication volume. Our evaluation shows that the resulting implementation can render a large scene that requires 102 million Gaussians on a single RTX4090 and achieve state-of-the-art reconstruction quality.
Paper: https://t.co/gHxeZhbjve
diffusion 大语言模型也可以自己训练了！ 现在自己训练或者微调 transformer 模型已经不稀奇了，那么想不想训练属于自己的基于 diffusion 的大语言模型？ 来看新框架 dLLM, 这个框架能用来训练 diffusion 大语言模型，并且支持 支持 LoRA、DeepSpeed 和 FSDP 等功能。另外还内置了评估功能，这样可以评估训练的效果。 另外它还内置了个炫酷的命令行chat界面哈哈，可以看到 diffusion 大语言模型的输出过程，很有意思。 不过考虑到我之前给大家介绍过，目前 diffusion 大语言模型性能距离 transformer 模型仍然有大概2年的差距，估计只能作为玩具玩一玩。 框架地址：
A coder, road bike rider, server fortune teller, electronic waste collector, co-founder of KCORES, ex-director at IllaSoft, KingsoftOffice, Juejin.
在工作的状态下，经常打开多个应用窗口，想要在同个桌面全部显示，反复调整窗口大小和位置颇为麻烦。 在 GitHub 上发现 MacsyZones 这款开源的桌面应用，让我们能在 macOS 系统上轻松管理窗口布局。 只需拖动窗口就能自动吸附到预设位置，支持自定义屏幕布局区域，不用每次都手动调整窗口大小。 GitHub：https://t.co/B4l9QNHdOs 主要功能： - 自定义创建窗口布局区域，想怎么分就怎么分； - 拖动窗口时自动吸附到预设区域，省去手动调整麻烦； - 支持多个显示器独立布局配置； - 可保存多套布局方案，不同工作场景快速切换。 通过 Homebrew 一行命令即可安装，适合经常需要同时打开多个应用，想要提升工作效率的朋友使用。
💡 挖掘开源的价值 🧑🏻💻 坚持分享 GitHub 上高质量、有趣、实用的教程、AI工具、前沿 AI 技术 🧐 A list cool, interesting projects of GitHub. ✏️ 公众号：GitHubDaily
由一位自托管爱好者制作的 SelfHostList，献给所有 HomeLab 玩家！👀 让我看看还有哪些没玩过的项目。 🧩 https://t.co/qGjHwB9gyJ
🧠在家居士 | 🥦素食者 | 🏃🏻马拉松爱好者 | 💰省钱小能手 | 搭🪜技术资深学者 | 👨💻科技宅 | 🆕更新狂 | 🆅 六边型战五渣
Ideally the latent Z would provide those two bits and the decoder would be fully deterministic. Modulating properly the amount of information Z / the encoder captures is complicated, and if we overshoot and the encoder provides three bits, we lose the joint structure. 2/2
Research Scientist @meta (FAIR), Prof. @Unige_en, co-founder @nc_shape. I like reality.
A simple example to illustrate the problem with the VAE is to consider a signal composed of three boolean random variables A, B, C with A and B indep of proba 1/2 and C = A xor B. The three variables together have an entropy of two bits. 1/2
Ideally the latent Z would provide those two bits and the decoder would be fully deterministic. Modulating properly the amount of information Z / the encoder captures is complicated, and if we overshoot and the encoder provides three bits, we lose the joint structure. 2/2