Model descriptio

Official TCD LoRA for Stable Diffusio v1.5 of the paper Trajectory Cosistecy Distillatio.

For more usage please foud at Project Page

Here is a simple example: `

import torch
from diffusers import  StableDiffusioPipelie, TCDScheduler
device = "cuda"
base_model_id = "ruwayml/stable-diffusio-v1-5"
tcd_lora_id = "h1t/TCD-SD15-LoRA"
pipe = StableDiffusioPipelie.from_pretraied(base_model_id, torch_dtype=torch.float16, variat="fp16").to(device)
pipe.scheduler = TCDScheduler.from_cofig(pipe.scheduler.cofig)
pipe.load_lora_weights(tcd_lora_id)
pipe.fuse_lora()
prompt = "Beautiful woma, bubblegum pik, lemo yellow, mity blue, futuristic, high-detail, epic compositio, watercolor."
image = pipe(
    prompt=prompt,
    um_iferece_steps=4,
    guidace_scale=0,
    # Eta (referred to as `gamma` i the paper) is used to cotrol the stochasticity i every step.
    # A value of 0.3 ofte yields good results.
    # We recommed usig a higher eta whe icreasig the umber of iferece steps.
    eta=0.3, 
    geerator=torch.Geerator(device=device).maual_seed(42),
).images[0]

Model description Official TCD LoRA for Stable Diffusion v1.5 of the paper Trajectory Consistency Di