lint

Signed-off-by: Vladimir Mandic <mandic00@live.com>

modules/seedvr/src/__init__.py

@@ -7,7 +7,7 @@ Version: 2.0.0 - Modular
 
 Available Modules:
 - utils: Download and path utilities
-- optimization: Memory, performance, and compatibility optimizations  
+- optimization: Memory, performance, and compatibility optimizations
 - core: Model management and generation pipeline (NEW)
 - processing: Video and tensor processing (coming next)
 - interfaces: ComfyUI integration
@@ -17,7 +17,7 @@ Available Modules:
 MODULES_AVAILABLE = {
     'downloads': True,          # ✅ Module 1 - Downloads and model management
     'memory_manager': True,     # ✅ Module 2 - Memory optimization
-    'performance': True,        # ✅ Module 3 - Performance optimizations  
+    'performance': True,        # ✅ Module 3 - Performance optimizations
     'compatibility': True,      # ✅ Module 4 - FP8/FP16 compatibility
     'model_manager': True,      # ✅ Module 5 - Model configuration and loading
     'generation': True,         # ✅ Module 6 - Generation loop and inference

modules/seedvr/src/common/config.py

@@ -71,15 +71,15 @@ def resolve_inheritance(config: Union[DictConfig, ListConfig]) -> Any:
 def import_item(path: Union[str, List[str]], name: str) -> Any:
     """
     Import a python item with fallback support.
-    
+
     Args:
         path: Single path string or list of paths to try (fallback order)
         name: Class/function name to import
-        
+
     Returns:
         Imported object
-        
-    Example: 
+
+    Example:
         import_item("path.to.file", "MyClass") -> MyClass
         import_item(["path1.to.file", "path2.to.file"], "MyClass") -> MyClass (first working path)
     """

modules/seedvr/src/common/seed.py

@@ -27,4 +27,3 @@ def set_seed(seed: Optional[int], same_across_ranks: bool = False):
         random.seed(seed)
         np.random.seed(seed)
         torch.manual_seed(seed)
-

modules/seedvr/src/core/__init__.py

@@ -29,13 +29,13 @@ from .infer import VideoDiffusionInfer
 __all__ = [
     # Model management
     'configure_runner',
-    'load_quantized_state_dict', 
+    'load_quantized_state_dict',
     'configure_dit_model_inference',
     'configure_vae_model_inference',
-    
+
     # Generation logic
     'generation_step',
-    'generation_loop', 
+    'generation_loop',
     'cut_videos',
     'prepare_video_transforms',
     'load_text_embeddings',
@@ -43,5 +43,5 @@ __all__ = [
 
     # Infer
     'VideoDiffusionInfer'
-] 
+]
 '''

modules/seedvr/src/core/generation.py

@@ -12,16 +12,16 @@ from src.utils.color_fix import wavelet_reconstruction
 def generation_step(runner, text_embeds_dict, cond_latents, temporal_overlap, device):
     """
     Execute a single generation step with adaptive dtype handling
-    
+
     Args:
         runner: VideoDiffusionInfer instance
         text_embeds_dict (dict): Text embeddings for positive and negative prompts
         cond_latents (list): Conditional latents for generation
         temporal_overlap (int): Number of frames for temporal overlap
-        
+
     Returns:
         tuple: (samples, last_latents) for potential temporal continuation
-        
+
     Features:
         - Adaptive dtype detection (FP8/FP16/BFloat16)
         - Optimal autocast configuration for each model type
@@ -111,7 +111,7 @@ def cut_videos(videos):
 def generation_loop(runner, images, cfg_scale=1.0, seed=666, res_w=720, batch_size=90, temporal_overlap=0, progress_callback=None, device:str='cpu'):
     """
     Main generation loop with context-aware temporal processing
-    
+
     Args:
         runner: VideoDiffusionInfer instance
         images (torch.Tensor): Input images for upscaling
@@ -121,10 +121,10 @@ def generation_loop(runner, images, cfg_scale=1.0, seed=666, res_w=720, batch_si
         batch_size (int): Batch size for processing
         temporal_overlap (int): Frames for temporal continuity
         progress_callback (callable): Optional callback for progress reporting
-        
+
     Returns:
         torch.Tensor: Generated video frames
-        
+
     Features:
         - Context-aware generation with temporal overlap
         - Adaptive dtype pipeline (FP8/FP16/BFloat16)
@@ -300,13 +300,13 @@ def generation_loop(runner, images, cfg_scale=1.0, seed=666, res_w=720, batch_si
 def prepare_video_transforms(res_w):
     """
     Prepare optimized video transformation pipeline
-    
+
     Args:
         res_w (int): Target resolution width
-        
+
     Returns:
         Compose: Configured transformation pipeline
-        
+
     Features:
         - Resolution-aware upscaling (no downsampling)
         - Proper normalization for model compatibility
@@ -328,15 +328,15 @@ def prepare_video_transforms(res_w):
 def calculate_optimal_batch_params(total_frames, batch_size, temporal_overlap):
     """
     Calculate optimal batch processing parameters
-    
+
     Args:
         total_frames (int): Total number of frames
         batch_size (int): Desired batch size
         temporal_overlap (int): Temporal overlap frames
-        
+
     Returns:
         dict: Optimized parameters and recommendations
-        
+
     Features:
         - 4n+1 constraint optimization
         - Padding waste calculation

modules/seedvr/src/models/dit_v2/attention.py

@@ -90,4 +90,3 @@ class FlashAttentionVarlen(nn.Module):
             return flash_attn_varlen_func(*args, **kwargs)
         except ImportError:
             return pytorch_varlen_attention(*args, **kwargs)
-

modules/seedvr/src/optimization/memory_manager.py

@@ -16,7 +16,7 @@ from src.models.dit_v2.rope import RotaryEmbeddingBase
 def preinitialize_rope_cache(runner) -> None:
     """
     🚀 Pre-initialize RoPE cache to avoid OOM at first launch
-    
+
     Args:
         runner: The model runner containing DiT and VAE models
     """
@@ -93,7 +93,7 @@ def preinitialize_rope_cache(runner) -> None:
 def clear_rope_cache(runner) -> None:
     """
     🧹 Clear RoPE cache to free VRAM
-    
+
     Args:
         runner: The model runner containing the cache
     """

modules/seedvr/src/optimization/performance.py

@@ -13,16 +13,16 @@ def optimized_video_rearrange(video_tensors: List[torch.Tensor]) -> List[torch.T
     """
     🚀 OPTIMIZED version of video rearrangement
     Replaces slow loops with vectorized operations
-    
+
     Transforms:
-    - 3D: c h w -> t c h w (with t=1)  
+    - 3D: c h w -> t c h w (with t=1)
     - 4D: c t h w -> t c h w
-    
+
     Expected gains: 5-10x faster than naive loops
-    
+
     Args:
         video_tensors: List of video tensors to rearrange
-        
+
     Returns:
         List of rearranged tensors in t c h w format
     """
@@ -81,16 +81,16 @@ def optimized_single_video_rearrange(video: torch.Tensor) -> torch.Tensor:
     """
     🚀 OPTIMIZED version for single video tensor
     Replaces rearrange() with native PyTorch operations
-    
+
     Transforms:
     - 3D: c h w -> 1 c h w (add temporal dimension)
     - 4D: c t h w -> t c h w (permute dimensions)
-    
+
     Expected gains: 2-5x faster than rearrange()
-    
+
     Args:
         video: Input video tensor
-        
+
     Returns:
         Rearranged tensor with temporal dimension first
     """
@@ -106,16 +106,16 @@ def optimized_sample_to_image_format(sample: torch.Tensor) -> torch.Tensor:
     """
     🚀 OPTIMIZED version to convert sample to image format
     Replaces rearrange() with native PyTorch operations
-    
+
     Transforms:
     - 3D: c h w -> 1 h w c (add temporal dimension + permute to image format)
     - 4D: t c h w -> t h w c (permute to image format)
-    
+
     Expected gains: 2-5x faster than rearrange()
-    
+
     Args:
         sample: Input sample tensor
-        
+
     Returns:
         Tensor in image format (channels last)
     """
@@ -130,12 +130,12 @@ def optimized_sample_to_image_format(sample: torch.Tensor) -> torch.Tensor:
 def temporal_latent_blending(latents1: torch.Tensor, latents2: torch.Tensor, blend_frames: int) -> torch.Tensor:
     """
     🎨 Temporal blending in latent space to avoid discontinuities
-    
+
     Args:
         latents1: Latents from previous batch (end frames)
         latents2: Latents from current batch (start frames)
         blend_frames: Number of frames to blend
-    
+
     Returns:
         Blended latents for smooth transition
     """
@@ -155,6 +155,3 @@ def temporal_latent_blending(latents1: torch.Tensor, latents2: torch.Tensor, ble
     blended_latents = weights1 * latents1 + weights2 * latents2
 
     return blended_latents
-
-
-