Add LTX-2 Support (#644)

* WIP, adding support for LTX2

* Training on images working

* Fix loading comfy models

* Handle converting and deconverting lora so it matches original format

* Reworked ui to habdle ltx and propert dataset default overwriting.

* Update the way lokr saves to it is more compatable with comfy

* Audio loading and synchronization/resampling is working

* Add audio to training. Does it work? Maybe, still testing.

* Fixed fps default issue for sound

* Have ui set fps for accurate audio mapping on ltx

* Added audio procession options to the ui for ltx

* Clean up requirements

toolkit/config_modules.py

@@ -207,6 +207,9 @@ class NetworkConfig:
         # -1 automatically finds the largest factor
         self.lokr_factor = kwargs.get('lokr_factor', -1)
         
+        # Use the old lokr format
+        self.old_lokr_format = kwargs.get('old_lokr_format', False)
+        
         # for multi stage models
         self.split_multistage_loras = kwargs.get('split_multistage_loras', True)
         
@@ -672,6 +675,9 @@ class ModelConfig:
         # kwargs to pass to the model
         self.model_kwargs = kwargs.get("model_kwargs", {})
         
+        # model paths for models that support it
+        self.model_paths = kwargs.get("model_paths", {})
+        
         # allow frontend to pass arch with a color like arch:tag
         # but remove the tag
         if self.arch is not None:
@@ -956,7 +962,7 @@ class DatasetConfig:
         # it will select a random start frame and pull the frames at the given fps
         # this could have various issues with shorter videos and videos with variable fps
         # I recommend trimming your videos to the desired length and using shrink_video_to_frames(default)
-        self.fps: int = kwargs.get('fps', 16)
+        self.fps: int = kwargs.get('fps', 24)
         
         # debug the frame count and frame selection. You dont need this. It is for debugging.
         self.debug: bool = kwargs.get('debug', False)
@@ -972,6 +978,9 @@ class DatasetConfig:
         self.fast_image_size: bool = kwargs.get('fast_image_size', False)
         
         self.do_i2v: bool = kwargs.get('do_i2v', True)  # do image to video on models that are both t2i and i2v capable
+        self.do_audio: bool = kwargs.get('do_audio', False) # load audio from video files for models that support it
+        self.audio_preserve_pitch: bool = kwargs.get('audio_preserve_pitch', False) # preserve pitch when stretching audio to fit num_frames
+        self.audio_normalize: bool = kwargs.get('audio_normalize', False) # normalize audio volume levels when loading
 
 
 def preprocess_dataset_raw_config(raw_config: List[dict]) -> List[dict]: