|
--- |
|
{} |
|
--- |
|
better version of the old huginn model, I THINK it's a little tamer now? less of a schizophrenic loose cannon, |
|
|
|
|
|
it's chronos, airoboros, hermes, beluga merged using my random-merge method, then merged with mythologic using model revolver, then merged with ledgerdemain, and the limarp lora. |
|
|
|
|
|
I'm aware that mythologic has some of the models huginn already had, but merging them in a second time seemed to smooth out some of it's weird behaviors, and then ledgerdemain allows it to do "step by step" thinking with character behaviors in roleplays. |
|
|
|
It is a little less verbose, unless you take some care in the prompt, where aledgedly it's a verbose as the old one, but it's a little harder to make it be as verbose. |
|
It benefits best from the alpaca format for instructions, but you can chat with it too. |
|
|
|
thanks to gryphe for helping with improving this merge and permitting me to add it to the official version! It means a lot! |
|
|
|
|
|
v1.3 will be actually trained on, I recently got access to some training servers thanks to the guanaco team! |