V0.2 updatedSingle character dataset with masked watermark. Trained to 2250 stepsThis is my very first attempt at making a LoRA for Flux Dev. Trained using <a target="_blank" rel="ugc" href="https://github.com/kijai/ComfyUI-FluxTrainer">https://github.com/kijai/ComfyUI-FluxTrainer</a>AdamW - Cosine with Restarts64 Dim64 NetworkLearning rate 3e-41600 steps (200 epoch) on 8 random images of jollyjack's art.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0a4a776f-f09a-4d8b-af4d-ea3332d28354/width=525/0a4a776f-f09a-4d8b-af4d-ea3332d28354.jpeg" />Seems to work best at 0.7 strength.

JollyJack_LoRA_Alpha

Changed the dataset to only be of single characters. Masked watermark. Trained to 2250 Steps (about 250 Epochs) I did a 3000 step version, but it was overtrained and kept putting heels on everyone.

Files with SHA256: d7630bcc2ccae7dac3cc69f195c93c51eef04d40d1d031670d3ad25e1011433d