Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
erinys 
posted an update 9 days ago
Post
1873
We shut down XetHub today after almost 2 years. What we learned from launching our Git-scaled product from scratch:
- Don't make me change my workflow
- Data inertia is real
- ML best practices are still evolving

Closing the door on our public product lets us focus on our new goal of scaling HF Hub's storage backend to improve devX for a larger community. We'd love to hear your thoughts on what experiences we can improve!

Read the full post: https://xethub.com/blog/shutting-down-xethub-learnings-and-takeaways

I read the article with great interest.

Git outta here! I’m not changing my workflow.
ML engineers and data scientists aren’t always software engineers

I believe this is an issue that HF currently faces as well.
Maybe the situation is a bit better in English-speaking countries, but in my country, HF is used because people are shown how to use it on external forums, blogs, social networking sites, and know-how sites, and few people are able to use HF on their own based on information provided by HF and its community.
Of course, there are many researchers who know a lot about coding, especially in the NLP field, but what if we expand the use cases to medicine, art, or music, for example? They don't make coding part of their job, so there are few people who know much about it.

Current HF is good for initial invention and experimentation by researchers, developers and their teams, but it is not as good as it could be for maintaining the ecosystem afterwards. The more the field moves into practical areas, the more people move away from HF.
I don't necessarily think that's a bad thing, and it's important to use different sites for different use cases.
However, I think there could be a little more ingenuity in HF to look over the AI resources of the entire net or society as a whole, and to facilitate the mutual transfer of information and resources.

Simply put, could there be more technical support, both manned and unmanned, for non-coders like painters, musicians, and doctors, or could there be more no-code, low-code, graphical usage? I think so.
Currently, even the official GUI tools are broken in places. People who can code don't notice it because it's not a problem for them. Neither do I. I think we should consider how critical it is that it actually goes unnoticed for so long. A geek-only service can easily fall apart.
There are so many questions on the forum but not enough respondents. There should be so many people out there...
HF is firmly established as an infrastructure for the entire generative AI world, but it is as a tool and service provider and storage, not as a community. At best it is just a temporary hub for inventors. I think it is even a daily progression.
HF is now a hub for things for coders and researchers, but it is not and is not likely to become a hub for information, people, culture, and civilization, and so on.

I hope what you have learned will be put to good use in HF. For a long time now, OSS and related projects tend to start smoothly because there is no leader, and end dumbly because there is no leader. It is difficult to manage a project by the wisdom of a group without relying on a superior individual, but I guess we have no choice but to do it.

HF is looking for customer requests, so if anyone has an opinion, please post it in the post below.
https://huggingface.co/posts/victor/964839563451127

·

This is great feedback @John6666 - and I've seen your suggestions in the other thread as well. As a non-ML engineer myself, it's been really interesting to explore HF with fresh eyes! We're doing some early exploration on HF understandability and discoverability in our team - would you be open to chatting sometime about potential approaches? We'd love to get your feedback!