Flux LoRA for creating whimsical illustrations. Use “illustration in the style of WHMSCPE001” to trigger the model. Best with DEV (works well even with fp8). To use with flux1-dev-bnb-nf4 try increasing the LoRA strength (say 1.3 and then adjust accordingly)
steps: 1000 lr: 0.0004 rank: 16 total number of training images: 17 (synthetic) Trained at Replicate using: https://replicate.com/ostris/flux-dev-lora-trainer/train
Weights for this model are also available in Safetensors format. Download them here: https://huggingface.co/bingbangboom/flux_whimscape/tree/main