You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Which level of error would you like to expose ?
For instance, trying to query the stabilityai/stable-diffusion space, I get a gradio internal error :
serialized_input = block.serialize(inputs[i])
AttributeError: 'Dataset' object has no attribute 'serialize'
But when querying the mirrored abidlabs/images space I might wait a long time to get no answer (understandable, with the current queue), or explicitely get an asyncio.exceptions.CancelledError much sooner.
The second might be caught and sent back to the channel as a timeout, but I don't know if users should be exposed to the first one as is.
In terms of the second error, I think we should also expose the queue ETA when people first make a request to a particular Space. However, this should probably be a separate issue. Would you like to go ahead and create it?
Instead of not replying, it would be better if errors were exposed
The text was updated successfully, but these errors were encountered: