I'm having a lot of fun with this new model.
☕ Support me on Ko-fi
Any contribution is appreciated and helps me keep working on open-source projects.
https://ko-fi.com/luisacaotica
☁️ Support through RunPod
You can also use my referral link when signing up for RunPod:
https://runpod.io?ref=kpjo0xmj
f you purchase RunPod credits through the referral, you get the computing resources you need, and I receive referral support as well. It's a win-win! It would also help me get access to more GPU resources so I can experiment with and train LoRAs alongside my other projects.
Description
No trigger, just go on.
FAQ
Comments (7)
Very nice model! Do you mind sharing your training config for this? In particular I'd like to know if you trained on base or directly on SFT.
lr: 0.001 weight decay: 0.05 diff guidance [yes] with 5.
trained on stf.
300 steps, but i started lower 0.0004 then i fixed with more 450 with 0.001
forgot to mention that is 4 dim 4 alpha
awesome lora. What is 8stepsdmd?