GitLab

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

* has read BaseRunner -> DefaultRunner -> WanRunner's init

* FirstBlock

* FirstBlock

* FirstBlock

* FirstBlock

* FirstBlock

* FirstBlock

* FirstBlock

* DualBlock

* DynamicBlock

* DynamicBlock

* DynamicBlock

* DynamicBlock

* DynamicBlock

* DualBlock Downsample Factor

* DualBlock Downsample Factor

* Downsample

* Downsample

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

* Enhance BaseRunner with model protocols and type hints and fix lora load

* Add support for multiple LoRAs configuration and enhance lora_strength argument handling

* doc: style fix