jy0205 / Pyramid-Flow

Code of Pyramidal Flow Matching for Efficient Video Generative Modeling

https://pyramid-flow.github.io/

MIT License

2.43k stars 238 forks source link

Guys, add Seed please. This is not serious to make tests with random seeds #147

Open ptits opened 3 weeks ago

ptits commented 3 weeks ago

def initialize_model_cached(variant, seed): key = variant torch.manual_seed(seed)

If you are using GPUs, set the seed for all GPUs

if torch.cuda.is_available():
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)  # for multi-GPU setups

# Check if the model is already in the cache

shot_241031_235815

feifeiobama commented 3 weeks ago

Thank you for the contribution! We will try to add the seed and other functionalities (e.g. text-to-image generation) in the new Gradio demo.

cocktailpeanut commented 1 week ago

@feifeiobama I've just sent a PR with the custom seed feature (the default is 0, which randomizes the seed value), as well as adding a way to switch back and forth between 384p and 768p, which automatically updates the maximum value of the duration slider between 16 and 31. Please check https://github.com/jy0205/Pyramid-Flow/pull/189

Giribot commented 6 days ago

Done ? (answer: yes !) (cf the new app.py) Maybe it"s time to close this topic ?