TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Do We Need a New Orchestration System for GPUs?

2 点作者 zenlikethat大约 2 年前

2 条评论

brucethemoose2大约 2 年前
There is something similar to what you described operating right now: Stable Horde: <a href="https:&#x2F;&#x2F;stablehorde.net&#x2F;" rel="nofollow">https:&#x2F;&#x2F;stablehorde.net&#x2F;</a><p>I&#x27;m not sure if it has the priority queueing system you mentioned, or how much it has generalized to non-SD loads, but it is functional.<p>Also, you should consider using a Huggingface diffusers UI&#x2F;backend instead of the automatic1111 UI, which is based on (and now inextricably linked to) the old Stability AI implementation. Maybe thats fine now, but its already a problem for optimizations people are cooking up (like Facebook&#x27;s AITemplate, torch.compile, SHARK&#x2F;MLIR for AMD GPUs, Intel OpenVINO...) which work with diffusers already.<p>I am bouncing between InvokeAI and VoltaML at the moment (as I cant get any of the optimizations to work in the auto1111 implementation anymore), but I can hardly keep up with everything on Github.
评论 #35464166 未加载
brianjking大约 2 年前
I&#x27;m only part of the way through the article, and yeah, interesting read.
评论 #35464156 未加载