Implements per-guide attention attenuation via log-space additive bias
in self-attention. Each guide reference tracks its own strength and
optional spatial mask in conditioning metadata (guide_attention_entries).
* Disable timestep embed compression when inpainting
Spatial inpainting not compatible with the compression
* Reduce crossattn peak VRAM
* LTX2: Refactor forward function for better VRAM efficiency
Jukka Seppänen
comfyanonymous
Tavi Halperin