Switching from one platform to another platform for model management

I have noticed that some practitioners start by using Hugging Face to manage their models, but later switch to GitHub, effectively abandoning Hugging Face. Could there be any specific reasons for this? I wonder if these practitioners initially train their models locally, publish them on Hugging Face for sharing and experimentation, and later move to GitHub for better management of commits and issues.
I could be mistaken, but I am genuinely curious. Does anyone have other ideas or insights about this?

1 Like

For example, the official libraries for Hugging Face itself are also maintained on github.
Hugging Face is a convenient platform for placing and training models, and for experimenting with and demonstrating model behavior, but if you want to use it for full-fledged software development, github is probably more convenient. If you want to use it for advertising models, X and other services are better.

The HF staff also generally use Hugging Face, github, and Discord separately. We users also often use other sites and Hugging Face together for convenience.

I don’t know whether this is the goal that Hugging Face has in mind or not, and the functions of Hugging Face itself may be enhanced, but as things stand, it is certainly convenient.

Thank you very much for your response. I really appreciate it. So, we can say that Hugging Face is an independent platform that can also do part of the GitHub activities.
Can I ask you something?
Do you think people train their model locally and then upload the trained model on Hugging Face before uploading the code and other implementation resources on GitHub OR do they upload the implementation resources on GitHub first and when they think everything is fine, they then upload the final model on Hugging Face?

What I can observe is that people do the same activities they perform on GitHub on Hugging Face as well. I might be wrong though.

1 Like

I think I’ve seen both patterns before. Both github and Hugging Face are platforms that you can use in whatever way is easiest for you when you come up with an idea.
The same goes for Google Colab, Lightning.ai, Kaggle, etc.

We can save money by Frankenstein-ing together various services!

1 Like