I’m facing an issue with AutoTrain using the Hugging Face graphical interface. Every time I start a training session, my Space automatically pauses within seconds, showing the message: “This Space has been paused by its owner”. However, I haven’t done anything to stop it.
I checked the logs, but they don’t clearly indicate where the issue is coming from. I also searched the forums and documentation but couldn’t find a solution.
Has anyone encountered this issue before or has any idea what might be causing it and how to fix it? Any help would be greatly appreciated!
It seems to be a kind of safety device to prevent the case where the program continues to run with a bug and consumes money, assuming that it is running on a paid Spaces. There is probably an error in the settings or something.
space pauses itself to save you money on failure or success. even after its paused, you can click on logs to see what happened. you can find the trained models in your huggingface account if the training finished.
I figured this out. It’s a setting you control when you first create your Space. At the bottom, there’s a setting called: PAUSE_ON_FAILURE. Set that to 0 instead of 1. Now your Space won’t pause when you encounter an error, and you’ll be able to check the relevant logs.
Sorry to ressurect this, but where’s this pause on failure button? Does this work for non AutoTrain apps? Mine is just a regular FastAPI app. There’s no setting in my space like that and my space randomly auto-pauses and gets into a pause/restart loop always refreshing the page; making it impossible to read the logs. I have no idea what happened. Any clues?
You can explicitly pause a Space via the HF API or GUI, but auto-pause is an AutoTrain feature, so it shouldn’t be present in a regular FastAPI app unless you consciously implement it… probably.