import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("baesik/kreamania_int8-convrot", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Kreamania Models — INT8 ConvRot Collection
Quantized INT8 versions of Kreamania (Krea 2) models using convert-to-quant. These models use significantly less VRAM and load natively in modern ComfyUI without extra nodes.
Conversion Command Used (Windows CMD)
ctq -i "C:\ComfyUI\models\diffusion_models\krea2\model_name.safetensors" --int8 --scaling_mode row --convrot --convrot-group-size 256 --comfy_quant --save-quant-metadata --krea2 --heur --low-memory --device cuda --manual-seed 42 --verbose VERBOSE --lr_patience 3 --lr_cooldown 2 --lr_shape_influence 0.5
- Downloads last month
- 151