Questions about model
Once it's uploaded, is it going to use linear scaling or is this more ntk alpha trained?
Any plans to Finetune on popular fanfiction databases or anything?
It'll be a quadruple merge, hermes-llongma + guanaco + limarp + storytelling, base model being hermes-llongma, which I don't know exactly how it was trained. I don't plan on doing any further finetuning on this model, but I plan on finetuning similar models in the future.
llongma is linear scaling context.
Guanaco seems like it would just degrade the model with redundant stuff and alignment?
Maybe, I have yet to test so it isn't absolute what I do. I found that it seems to make hermes more creative though.
I hope to try the model some day if you upload it.
Did the model not ending up panning out?
Been working on other stuff recently, but I am finishing this model now. Should be uploaded in a few hours.
Uploading right now.
Nice!