nyuuzyou

Thank you for your response and your work.

CC @Yntec @Lyte @John6666 @GhostGate

I'm waiting for at least an announcement explaining what's going on and at most how this quota is calculated

HOW

Update
https://huggingface.co/posts/nyuuzyou/189643732365912

You don't have to ask. You just continue to upload and as long your models provide value to the AI community, they will be fine with it.

That's not entirely true. For example, I had to request a limit increase for my anime art dataset containing NSFW content (which fully complies with HF rules). Here is the response I received:

While HF's policies don't explicitly state they won't provide storage and computational quota for NSFW content, my experience suggests otherwise. Does this dataset and potential models trained on it provide value to the AI community? I believe so. But would they grant a quota increase for it? You can see their answer above.

I'm glad to hear that. But asking for a quota to publish a 100MB dataset is a bit strange.

Anyway, until an official announcement is made, I'm going to act as normal. I have started backing up locally just in case...

I only hope that no model disappears in surprise when it really doesn't need to disappear.😓

So far, no restrictions have been implemented, and I can still upload new files, so the situation isn't critical yet.

I hope Hugging Face will find ways to expand storage for free or provide unlimited storage for PRO users instead of the current 1 TB limit. Kaggle offers unlimited storage for public repositories per account, with a limit of 200 GB per repository (which is less generous than Hugging Face's 300 GB). However, Kaggle won't come close to replacing Hugging Face's functionality, so we have nowhere else to turn.

We might need someone from hugging face to actually weigh in. Maybe someone can @ a member in here, if anyone knows.

I'm sure the team was aware of the community's reaction to these changes before they were implemented. Now we'll just have to wait for the full details to come out.

I hope that there will be some form of a contingency plans, otherwise most of the finetunes and quants on large models might need to be removed and I hope that won't happen. I would hate to have to go Q4 only from now on, just because no one can upload multiple quants because no space.

I hope Hugging Face will find ways to expand storage for free or provide unlimited storage for PRO users instead of the current 1 TB limit. Kaggle offers unlimited storage for public repositories per account, with a limit of 200 GB per repository (which is less generous than Hugging Face's 300 GB). However, Kaggle won't come close to replacing Hugging Face's functionality, so we have nowhere else to turn.

I've published almost 70 datasets, and from what I've seen, a combination of downloads and likes seems to be the way to go. My dataset nyuuzyou/subdomains has a few likes, but at its peak it had over 4,000 downloads in a month, and it wasn't in trending at all.