Implement Autoregressive DDPM

[marianovitasari20]

Problem / Goal

Currently, the DDPM is not autoregressive. Making it autoregressive will likely improve the current DDPM results within the main pipeline.

What does good look like?

• The autoregressive DDPM runs successfully without errors.
• Training and inference complete as expected.
• Results show measurable improvement over the current non-autoregressive DDPM.

Related context

No response