Turbocharge Diffusion Video Generation (Pro)

This tutorial demonstrates how to use the pruna package to optimize a video generation pipeline. We will use the HunyuanVideo model as an example. Any execution times given below are measured on a A10G GPU, as this tutorial requires at least 21GB of GPU memory.

1. Loading the Stable Diffusion Video Model

First, load your video generation model.

import torch
from diffusers import HunyuanVideoPipeline, HunyuanVideoTransformer3DModel
from diffusers.utils import export_to_video

model_id = "tencent/HunyuanVideo"
transformer = HunyuanVideoTransformer3DModel.from_pretrained(
    model_id,
    subfolder="transformer",
    torch_dtype=torch.bfloat16,
    revision="refs/pr/18",
    cache_dir="/efs/hf_cache",
)
pipe = HunyuanVideoPipeline.from_pretrained(
    model_id,
    transformer=transformer,
    torch_dtype=torch.float16,
    revision="refs/pr/18",
    cache_dir="/efs/hf_cache",
).to("cuda")

2. Initializing the Smash Config

Next, initialize the smash_config.

from pruna_pro import SmashConfig, smash

# Initialize the SmashConfig
smash_config = SmashConfig()

smash_config['cacher'] = 'adaptive'

3. Smashing the Model

Now, you can smash the model, which will take around 40 seconds. Don’t forget to replace the token by the one provided by PrunaAI.

# Smash the model
smashed_model = smash(
    model=pipe,
    token="<your_pruna_token>",
    smash_config=smash_config,
)

4. Running the Model

Finally, run the model to generate the video with accelerated inference.

output = pipe(
    prompt="A cat walks on the grass, realistic",
    height=256,
    width=256,
    num_frames=129,
    num_inference_steps=30,
    generator=torch.Generator("cuda").manual_seed(1)
).frames[0]
export_to_video(output, "hunyuan_video_smashed.mp4", fps=15)

Wrap Up

Congratulations! You have successfully smashed a stable diffusion video generation model. You can now use the pruna package to optimize any custom stable diffusion video generation model. The only parts that you should modify are step 1 and step 4 to fit your use case.