generate_samples module

generate_samples.generate_samples(diffusion_model, results_dir, ema_decay=0.995, ema_update_every=10, num_samples=100, batch_size=1, start_sample_idx=0)[source]

Generate samples using the given diffusion model and save them to the specified directory.

Parameters:

  • diffusion_model (GaussianDiffusion) – The diffusion model used for generating samples.

  • results_dir (str) – The directory where the generated samples will be saved.

  • ema_decay (float, optional) – The decay rate for the exponential moving average (EMA) of the diffusion model. Defaults to 0.995.

  • ema_update_every (int, optional) – The number of steps between EMA updates. Defaults to 10.

  • num_samples (int, optional) – The total number of samples to generate. Defaults to 100.

  • batch_size (int, optional) – The batch size used for generating samples. Defaults to 1.

  • start_sample_idx (int, optional) – The starting index for the generated sample filenames. Defaults to 0.