I made a heretic ara version (8/100 refusals, 0.079 KL)
#4
by kabachuha - opened
If anyone's wondering, there is an ARA-LoRA adapter I made for this model:
https://huggingface.co/kabachuha/G4-MeroMero-v2-31B-Heretic-ARA-LoRA
Quantized:
https://huggingface.co/kabachuha/G4-MeroMero-v2-31B-Heretic-ARA-LoRA-GGUF
To use in llama.cpp:
./llama-server <your current command> --lora-scaled "/media/kabachuha/doc/G4-MeroMero-v2-31B-Heretic-ARA-LoRA.gguf:1.0"
Oh thank you! I will try this tonight!
@Yingyaeliae It works quite great and more creative (than the older Gemma tunes) for Minimax prompt creation, and also short stories, so having fun with it now