Kartoffel-Deepfry-12B
mistral-nemo-kartoffel-12B finetuned on Schule-DPO
Method
QLoRA ORPO tuned with 1x RTX A6000 for 5 epochs. Rank 16 LoRA 32 alpha, 2e-4 learning rate cosine schedule.
mistral-nemo-kartoffel-12B finetuned on Schule-DPO
QLoRA ORPO tuned with 1x RTX A6000 for 5 epochs. Rank 16 LoRA 32 alpha, 2e-4 learning rate cosine schedule.