quick_inference.py

import torch
from diffusers import AnimateDiffPipeline, MotionAdapter, EulerDiscreteScheduler
from diffusers.utils import export_to_gif
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file


def inference(step,
              base="emilianJR/epiCRealism",
              prompt="A girl smiling"):
    
    device = "cuda"
    dtype = torch.float16

    # step = 4  # Options: [1,2,4,8]
    repo = "ByteDance/AnimateDiff-Lightning"
    ckpt = f"animatediff_lightning_{step}step_diffusers.safetensors"
    # base = "emilianJR/epiCRealism"  # Choose to your favorite base model.
    
    adapter = MotionAdapter().to(device, dtype)
    adapter.load_state_dict(load_file(hf_hub_download(repo, ckpt), device=device))
    pipe = AnimateDiffPipeline.from_pretrained(base, motion_adapter=adapter, torch_dtype=dtype).to(device)
    pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing", beta_schedule="linear")

    output = pipe(prompt=prompt, guidance_scale=1.0, num_inference_steps=step)
    export_to_gif(output.frames[0], "animation.gif")


if __name__ == "__main__":
    import argparse
    
    parser = argparse.ArgumentParser()
    
    parser.add_argument("--step", type=int, choices=[1,2,4,8], default=4)
    
    parser.add_argument("--base", type=str, default="emilianJR/epiCRealism")
    
    parser.add_argument("--prompt", type=str, default="A girl smiling")
    
    args = parser.parse_args()
    
    inference(args.step, args.base, args.prompt)