https://github.com/rqluo/mineai

Decentralized AI sharing services
https://github.com/rqluo/mineai

Last synced: 3 months ago
JSON representation

Decentralized AI sharing services

• Host: GitHub
• URL: https://github.com/rqluo/mineai
• Owner: RQLuo
• Created: 2024-08-09T17:13:54.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-08-24T03:25:36.000Z (almost 2 years ago)
• Last Synced: 2025-04-23T18:11:22.491Z (about 1 year ago)
• Language: HTML
• Size: 56.6 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# MixTeX-CommuniAI

Decentralized AI sharing services

这是我之前用于在线部署的网站，原理是将用户的图像发送到服务器，再由服务器发送到我的主机，由我的gpu推理后，将结果返还给服务器，再返还给用户。

![IMG_20240809_154524](https://github.com/user-attachments/assets/fe3e6112-3dce-4d2a-8d81-d1755ecc8b11)

我们的计划如图右侧目标，需要有专业的网络方面志愿者来完善，我希望将这个做成去中心化，将来的开源社区模型会越来越好，MixTeX也会在将来推出参数较大的问答模型，届时模型难以在本地用cpu推理，不过很多同学的轻薄本又没有gpu。

由于，MixTeX希望延续开源和共享精神，并且数学物理的问答与识别都是针对学生和科研工作者，我们不愿商业化污染这片净土。

因此，仅靠捐赠我们是无法承担高昂的高性能在线gpu服务，也就有了我们的去中心化服务的草图，我们希望有计算资源（至少GTX1060）的同学，能够做出一些贡献，将自己的电脑作为gpu计算节点，服务器会根据节点状态分发用户请求。

This was the website I previously used for online deployment. The principle is to send users' images to the server, which then forwards them to my host. My GPU performs inference on them, and the results are sent back to the server and then to the users.

Our plan, as shown in the target on the right side of the image, requires professional volunteers in the network domain to perfect it. We aim to make this decentralized. Future open-source community models will continue to improve, and MixTeX will eventually release larger parameter Q&A models. At that point, it will be difficult to run the models locally with a CPU, but many students have thin laptops without GPUs.

Because MixTeX wishes to continue the spirit of open source and sharing, and since mathematical and physical Q&A and recognition are aimed at students and researchers, we are reluctant to commercialize this pure field.

Therefore, relying solely on donations, we cannot afford the high costs of high-performance online GPU services. This led to the draft of our decentralized service. We hope that students with computing resources (at least a GTX 1060) can contribute by turning their computers into GPU computing nodes. The server will distribute user requests based on the status of these nodes.