GitLab

* [Fix] Fix audio model offload bug.

* Fix

* Fix

* Fix

* Fix

* Fix

* Fix

* bugfix:adapt to  5B dit model, derive attention_head_dim from config[dim]

* [Fix] Wan2.2 Vae Encode refactor: drop args parameter and use self.cpu_offload

refactor: streamline run_pipeline method in WanAudioRunner for improved clarity and efficiency, including enhanced error handling and memory management

refactor: replace optimize_latent_size_with_sp with get_optimal_patched_size_with_sp in WanAudioRunner for improved clarity and functionality

refactor: rename initialize_once to initialize in WanAudioRunner and update run_pipeline to reflect changes

refactor: optimize adaptive resizing logic in WanAudioRunner for improved aspect ratio handling and latent size calculation

refactor: update WanTransformerInfer and VideoGenerator to improve audio handling and progress tracking

refactor: update DefaultRunner to return audio as None and modify WanAudioRunner to handle audio saving and frame interpolation