Krea-2 LoRA — PixelCraft UI

LoRA (rank 32, alpha 32) train trên Krea-2 Raw bằng ai-toolkit / diffusion_trainer.

  • Trigger word: PixelCraft UI
  • Steps: 1000, lr 1e-4, optimizer adamw8bit, bf16
  • Resolutions: 768 / 1024 / 1280

Files

  • pytorch_lora_weights.safetensors — trọng số LoRA
  • config.yaml — config train gốc

Cách dùng

from diffusers import DiffusionPipeline
import torch

pipe = DiffusionPipeline.from_pretrained("krea/Krea-2", torch_dtype=torch.bfloat16).to("cuda")
pipe.load_lora_weights("kipgin/krea2_raw_768_102_1280_lora")
image = pipe("PixelCraft UI, a clean dashboard interface").images[0]
Downloads last month
6
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support