I'm using tabby worker::completion
to register a child worker with the master. However, when I start a second worker with the same configuration, it doesn’t seem to register successfully according to the cluster information. It appears that a master can only accept one model and one chat model? Even so, for my needs, I would expect the new model worker to successfully register and then have the related content load balanced onto this new model worker. But from what I’ve tested so far, it seems I must first kill the old model worker. Is this normal behavior?
moqi
Asked on Jan 18, 2024
Yes, there is a designed limitation – one chat worker plus one completion worker – for free tier of the enterprise license.