#I've been considering training my own
1 messages ยท Page 1 of 1 (latest)
u could have several mini agent models. and a big one for knowlege relatet stuff.
what do u think ? maybe it is possible to train the core logig of openclaw into a tiny model.
maybe several models for specific tasks.
the biggest problem here semms to get a good dataset how the big models just handle the tasks given by openclaw.or maybe i am totally wrong by my understanding how this works...