Cannot execute any model with my API Token, models are timed out

Inference Endpoints on the Hub
1

Hello to all the members of this amazing community! I’ve been on Hugging Face quite recently, so I apologize if this question is too dumb.

I have a small non-commercial Telegram bot that allows users to generate texts using a model that I have trained and choose the best of these texts. I wrote it just today and actively tested it alone during the day (so i have not users yet in fact). I uploaded my model to HF and use a free version of the API to generate texts.

Until some time
everything was fine, but now I can’t execute my own or any other model through the API. I get the invariable error “model time out”.

And this applies not only to the API, but also to the widget in the model cards - I can’t apply the model there either.

I absolutely do not allow a situation in which many requests would be made at the same time. Moreover, their number is not so large: I think I made about 50-60 requests that day.

Please tell me what could be the reason for such a reaction
to my requests? Are there any specific strict restrictions on the free API, for example, on the number of requests per day or on the interval between them?

I would appreciate your help!

2

I’m having the same problem, getting model timeouts for all of my API calls to my model for the past 1-2 hours. Not sure what is causing this.

Edit: just to be clear, I am using the free Inference API, not the paid Inference Endpoints

1 Like
3

It just started working again with my token, hooray! I would still like to find out if it was an HF failure or a limitation of a particular user.

1 Like
4

A few days later, it started happening again. Still, I would like to get an answer: are these user restrictions or unavoidable failures of the free API? I understand perfectly well that you cannot and should not guarantee trouble-free operation on a free plan and I am very grateful that it exists in principle, but there is just a feeling that everything works with other tokens, and therefore it looks like a restriction of a specific user…

5

hey @Futyn-Maker did you find the reason behind the time-outs ?

1 Like
6

No, at the time when this was relevant, it was fixed for me - after those two cases, timeouts no longer occurred. At the moment, the problem is simply irrelevant for me, because there is no longer free inference for unpopular models, as far as I know.

1 Like