Add tiny Wan debug pipeline

54c3b8d verified 8 days ago

1.01 kB

license: apache-2.0
library_name: diffusers
pipeline_tag: text-to-video
tags:
  - diffusers
  - wan
  - tiny-random
  - debug

Tiny FastWan2.1 T2V DMD Debug Pipeline

This is a randomly initialized, tiny WanDMDPipeline fixture for FastVideo/FastWan2.1-T2V-1.3B-Diffusers. FastVideo FastWan2.1 DMD-style text-to-video layout represented as a Diffusers-format artifact with _class_name patched to WanDMDPipeline for FastVideo VideoGenerator load-path debugging.

It is intended for fast load-path and inference-control debugging only. It is not trained and should not be used for generation quality evaluation.

import os
from fastvideo import VideoGenerator

os.environ["FASTVIDEO_ATTENTION_BACKEND"] = "TORCH_SDPA"
generator = VideoGenerator.from_pretrained(
    "Erland/tiny-fastwan2.1-t2v-dmd-debug",
    num_gpus=1,
)
try:
    generator.generate_video(
        "debug prompt",
        output_path="my_videos/",
        save_video=True,
    )
finally:
    generator.shutdown()