Hello, upon reading your article and checking the code, it seems to me that the Occupancy-aware fusion is missing.
control = control_model(x=x_t, hint=hint, [timesteps=t)]
control_scales = [1.0] * len(control)
control = [c * scale for c, scale in zip(control, control_scales)]
model_output = model(x_t, self._scale_timesteps(t), control, **model_kwargs
I might be looking in the completely wrong place. In that case, can you help me and point me towards it
Hello, upon reading your article and checking the code, it seems to me that the Occupancy-aware fusion is missing.
I might be looking in the completely wrong place. In that case, can you help me and point me towards it