U-Net Architecture and Conditioning

Implement sinusoidal positional encodings for timestep conditioning and integrate them into the U-Net

Integrate self-attention and cross-attention mechanisms to handle spatial dependencies and external conditioning

Implement class conditioning using learned embeddings

Explain the mechanism of classifier guidance and implement classifier-free guidance (CFG)

Tune CFG scale and diagnose common artifacts arising from guidance

Analyze cross-attention maps to debug prompt alignment failures

Apply gradient checkpointing to reduce memory usage during training