Update api/ltx/ltx_aduc_pipeline.py

api/ltx/ltx_aduc_pipeline.py

@@ -133,9 +133,8 @@ class LtxAducPipeline:
         
         temp_latent_paths = []
         overlap_condition_item: Optional[LatentConditioningItem] = None
-
         current_conditions = initial_conditions
-        
+
         try:
             for i, chunk_prompt in enumerate(prompt_list):
                 logging.info(f"Processing scene {i+1}/{num_chunks}: '{chunk_prompt[:50]}...'")
@@ -143,15 +142,15 @@ class LtxAducPipeline:
                 current_frames_base = frames_per_chunk if i < num_chunks - 1 else total_frames - ((num_chunks - 1) * frames_per_chunk)
                 current_frames = current_frames_base + (overlap_frames if i > 0 else 0)
                 current_frames = self._align(current_frames, alignment_rule='n*8+1')
-
-                current_conditions = initial_conditions if i == 0 else []
-                if overlap_condition_item: current_conditions.append(overlap_condition_item)
-
-                chunk_latents = self._generate_single_chunk_low(
-                    prompt_x=chunk_prompt, num_frames_x=current_frames, seed_x=used_seed,
-                    conditioning_items_x=current_conditions, 
-                    **kwargs
-                )
+               
+                kwargs.pop("prompt", None) 
+                kwargs.pop("num_frames", None) 
+                kwargs.pop("seed", None) 
+                
+                kwargs["prompt"] = chunk_prompt
+                kwargs["num_frames"] = current_frames
+                
+                chunk_latents = self._generate_single_chunk_low(**kwargs)
                 if chunk_latents is None: raise RuntimeError(f"Failed to generate latents for scene {i+1}.")
 
                 if is_narrative and i < num_chunks - 1:
@@ -161,8 +160,11 @@ class LtxAducPipeline:
                         media_frame_number=0,
                         conditioning_strength=1.0
                     )
-                    current_conditions=overlap_condition_item
-                
+                    kwargs.pop("conditioning_items", None) 
+                    kwargs["conditioning_items"] = overlap_condition_item
+                else:
+                    kwargs.pop("conditioning_items", None) 
+                    
                 if i > 0: chunk_latents = chunk_latents[:, :, overlap_frames:, :, :]
                 
                 chunk_path = RESULTS_DIR / f"temp_chunk_{i}_{used_seed}.pt"
@@ -221,8 +223,6 @@ class LtxAducPipeline:
     
     @log_function_io
     def _generate_single_chunk_low(
-        prompt_x:str, num_frames_x:int, seed_x:int,
-        conditioning_items_x:LatentConditioningItem,
         **kwargs
     ) -> Optional[torch.Tensor]:
         """[WORKER] Calls the pipeline to generate a single chunk of latents."""
@@ -234,13 +234,13 @@ class LtxAducPipeline:
 
         call_kwargs = {
             "cfg_star_rescale": "true",
-            "prompt": prompt_x,
+            "prompt": kwargs["prompt"],
             "negative_prompt": kwargs['negative_prompt'],
             "height": downscaled_height,
             "width": downscaled_width,
-            "num_frames": num_frames_x,
+            "num_frames": kwargs["num_frames_x"],
             "frame_rate": int(DEFAULT_FPS),
-            "generator": torch.Generator(device=self.main_device).manual_seed(seed_x),
+            "generator": torch.Generator(device=self.main_device).manual_seed(kwargs['seed_x']),
             "output_type": "latent",
             "media_items": None,
             "decode_timestep": self.config["decode_timestep"],
@@ -260,7 +260,7 @@ class LtxAducPipeline:
         call_kwargs.update(first_pass_config)
         ltx_configs_override = kwargs.get("ltx_configs_override", {}).copy()
         call_kwargs.update(ltx_configs_override)      
-        call_kwargs['conditioning_items'] = conditioning_items_x
+        call_kwargs['conditioning_items'] = kwargs["conditioning_items"]
         
         with torch.autocast(device_type=self.main_device.type, dtype=self.runtime_autocast_dtype, enabled="cuda" in self.main_device.type):
             latents_raw = self.pipeline(**call_kwargs).images