Inference API Issues


My team have currently been using the Inference API under the Lab pricing model specifically to use the philschmid/bart-large-cnn-samsum model to summarise conversations. The API and model were working fine up until a few days ago when we started receiving the following error every-time we try and call the model:

error_message 'unknown error' exception HuggingFaceModelNotAvailableYetError('unknown error') response <Response [500]> response_content { error: 'unknown error', warnings: [ "forward() got an unexpected keyword argument 'early_stopping_sada'" ] }

This is very strange because we hadn’t changed anything in the model itself. This morning I was experimenting with feeding different inputs to the model locally though such as early_stopping and number_of_beams and had seen a similar error message locally on my machine.

Whilst debugging this issue we found that somehow our model got upinned from our organization on the hugging face hub and so to fix this we repinned the model and are no longer seeing this error. We want to get to the root cause of why this might have happened though and we’re hoping you might be able to help us with this.

We also constantly keep seeing our model as unpinned on our organisations page on the hub.

Also apologies if this is not the right forum. We were note given an email to reach despite having that a feature in the Inference pricing model we chose.

Many thanks in advance,