Increased and improved dataset
Increased training steps
Used improved prompts (now Vakula will be generated with mustaches by default as should be)
Added papakha/kubanka hat
Reduced LoRA size (by using different train methods)
Increased and improved dataset
Increased training steps
Used improved prompts (now Vakula will be generated with mustaches by default as should be)
Added papakha/kubanka hat
Reduced LoRA size (by using different train methods)