Wan-AI/Wan2.1-T2V-1.3B-Diffusers output noise

[jiqing-feng]

Describe the bug

Script copied from https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B-Diffusers

import torch
from diffusers import AutoencoderKLWan, WanPipeline
from diffusers.utils import export_to_video

# Available models: Wan-AI/Wan2.1-T2V-14B-Diffusers, Wan-AI/Wan2.1-T2V-1.3B-Diffusers
model_id = "Wan-AI/Wan2.1-T2V-1.3B-Diffusers"
vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=torch.float32)
pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
pipe.to("cuda")

prompt = "A cat walks on the grass, realistic"
negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"

output = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    height=480,
    width=832,
    num_frames=81,
    guidance_scale=5.0
).frames[0]
export_to_video(output, "output.mp4", fps=15)

Reproduction

The output is noise

Logs

output like this:

System Info

diffusers 0.37.0.dev0
transformers 5.2.0.dev0
torch 2.10.0

Who can help?

@DN6 @sayakpaul

[sayakpaul]

Do you have a version that you knew to be working?

[asomoza]

seems this happens with when transformers > 4.57.3 is installed (so all the 5 versions)

[jiqing-feng]

Fixed in huggingface/transformers#43880