Argument for seed?

[SoftologyPro]

So the results can be reproduced if need be.

[a43992899]

Will support. See todos.

[alisson-anjos]

def set_seed(seed=42):
    torch.manual_seed(seed)  # Sets seed for PyTorch operations
    torch.cuda.manual_seed_all(seed)  # Sets seed for all CUDA GPUs (if available)
    np.random.seed(seed)  # Sets seed for NumPy
    random.seed(seed)  # Sets seed for Python's built-in random module
    torch.backends.cudnn.deterministic = True  # Ensures deterministic results
    torch.backends.cudnn.benchmark = False  # Disables benchmarking for consistency

https://github.com/alisson-anjos/YuE-Interface/blob/main/inference/infer.py

[hf-lin]

We have updated inference script to support this.