LitDDIM

class dmme.lit_modules.LitDDIM(lr: float = 0.0002, warmup: int = 5000, decay: float = 0.9999, diffusion_model: Optional[DDIM] = None, model: Optional[Module] = None, timesteps: int = 1000, sample_steps: int = 50, tau_schedule: str = 'quadratic')

LightningModule for accelerated sampling with DDIM using LitDDPM’s checkpoints

Parameters:

• lr – learning rate, defaults to 2e-4

• warmup – linearly increases learning rate for warmup steps until lr is reached, defaults to 5000

• decay – EMA decay value

• diffusion_model – overrides diffusion_model DDIM

• model – overrides model passed to DDIM

• timesteps – default timesteps passed to DDIM

• sample_steps – default sample steps passed to DDIM

• tau_schedule – default tau schedule passed to DDIM