Sloppy but somewhat effective LoRa model of Paris Dylan. Training set is basic, 5 (carefully selected) images over 35 epochs (175 steps) utilising the Prodigy optimizer. Simply don't have the time to be training 4000+ steps only to find out the results are trash at this time.
Model is most effective at around 1.25 - 1.35 weighting, suggesting there is room for improvement. As of now though this will suffice as a decent base starting point for potential future revisions.
This LoRa has a tendency to perform well with 'cute' checkpoints, have only briefly tested but realcartoonRealistic, artUniverse, dreamshaper_8 and colorful_v30 all yield relatively successful outputs along with most anime checkpoints. Absolute_Reality_v16 is a good starting point for more realistic outputs.
Keep your prompts simple, just add Happy or Big Smile to make the likeness pop when producing images. The model should remain fairly flexible even at higher weights, if not then simplify your prompts or lower the weights.
Enjoy!
描述:
v.0.1
训练词语:
名称: ParisD.safetensors
大小 (KB): 73841
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
