v-gen-ai/qwen-calibri

Paper: Calibri: Enhancing Diffusion Transformers via Parameter-Efficient Calibration

Calibri Qwen Image

Guide to run:

import torch
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(
    "makriot/qwen-calibri", 
    custom_pipeline="makriot/qwen-calibri",
    torch_dtype=torch.bfloat16,
    trust_remote_code=True
)
pipe.to("cuda")

image = pipe(
    prompt="A futuristic cityscape", 
    true_cfg_scale=4.0, 
    num_inference_steps=15,
    height=512,
    width=512,
).images[0]

Model was trained with resolution 512 x 512, but it is possible to run with other resolutions.