Hi everyone,
I’m Célina, one of the maintainers of the huggingface_hub library.
Really sorry for the inconvenience! The issue with HF Inference API should be fixed now.
Don’t hesitate to report any other issues you’re experiencing with HF Inference API or Inference Providers in general and of course, we welcome any feedback from the community 
Again, we’re really sorry for this!
All works, thank you!
already working?
Thanks for the quick fix @celinah
Hello @celinah - we are still getting the same issue - 404. We are using custom API keys and using together.ai as the provider
I’m getting this message in my log:
huggingface_hub.errors.HfHubHTTPError: 503 Server Error: Service Temporarily Unavailable for url: https://router.huggingface.co/hf-inference/models/google/gemma-2-2b-it
same goes for gemma-2-9b-it. 2b worked yesterday. If I follow the link, I get:
404
Sorry, we can’t find the page you are looking for.
if I try gemma 2-2b-it directly on the model page, it get this message:
Error forwarded from backend: Request failed during generation: Server error: error trying to connect: No such file or directory (os error 2)
If I ask again (in my free huggingface space app) I now get: An error occurred: 402 Client Error: Payment Required for url: https://router.huggingface.co/hf-inference/models/google/gemma-2-2b-it (Request ID: Root=1-67fcc32d-4f841d02600097901062b2a5;aaa2b4be-79ba-44e3-b920-12c18f76f629)
You have exceeded your monthly included credits for Inference Providers. Subscribe to PRO to get 20x more monthly included credits.
same now on the model page (which..makes no sense?):
This is probably an error in the whole Hugging Face system. I think it’s something like a setting error.
I think (I hope) that HF has already noticed and is working on the recovery without the need for a report this time…
If you do need to report, these github issues or HF Discord are the quicker way to do it.
Issue seems to be back again
I’m having a similar issue, but with image-to-image tasks. I keep getting the same error message (Model lambdalabs/sd-image-variations-diffusers inference is not supported HF inference api), regardless of the model I try
We are getting 404 error on these models
{“error”:“Model nvidia/cycle-diffusion does not exist”} Status: 404
{“error”:“Model yisol/IDM-VTON inference is not supported HF inference api”} Status: 404
Same, HF is a mess. This happens too often
far far too often.
i try to be understanding about this *, i mean
i feel like this says alot for expectations we should have,
but with media like this from HF it sets the expectations elsewhere
“The mission of Hugging Face is to democratize good machine learning and maximize its positive impact across industries and society. Not only do we strive to advance Transformer models, but we also work hard on simplifying their adoption.”
we cant hold them to their words, obviously, i laugh thinking just for one second that we might could, it all comes down to money. BUT there are better ways to get there other than ripping off potential customers by means of 404/503 daily.
this is a CLASSIC case of BAIT N SWITCH. a pure lack of ethics and conscious
all i can say is Google COLAB !!!
i dont even know WHY they have a single * concern about replacing some unseen ticker symbol with a huggy face when,
notice the .. contradiction already in their speak versus their actions?
open source? democritization? yet TRADED? who are their investors then if not the open public they claim to huggy face with ?
jesus christ, they dont even need us, let alone strive to offer the world access.
they’ve shut off stock access, they’ve even shut off SD access,
and they didnt do it because its about MONEY, they HAVE the biggest investors out there
so why bait n switch? its not about the money its about #s, and TRAFFIC, and amount of USERS. Thats what they investors are interested, THAT and feeding their own MODELS (like github / MS do) . so investors benefit
and HF benefits through private US trades FOR, lower to no corporate income taxes, Fewer restrictions on profit repatriation, Fewer obligations to disclose UBOs (ultimate beneficial owners), which becomes a security exchange concern
anyhow, thats off topic. we should just focus on 404/503 rates vs HF IF api rates
but after seeing their investor list, its all a mute point anyhow.
we have no say, and never where a concern for the company from the git go
It seems that major models such as FLUX and SD3.5 have returned to working condition. LoRA is still not working, but overall I feel that it has recovered to a state similar to a few days ago.
