torch 2.6.0 support

Signed-off-by: Vladimir Mandic <[email protected]>

CHANGELOG.md

@@ -4,31 +4,33 @@
 
 NYE refresh release with quite a few optimizatios and bug fixes...
 
-- **LoRA**:
+- **LoRA**:  
   - **Sana** support  
   - quantized models support  
   - fuse support with on-demand apply/unapply  
   - add legacy option in *settings -> networks*  
-- **HunyuanVideo**
+- **HunyuanVideo**:  
   - optimizations: full offload, quantization and tiling support  
-- **LTXVideo**
+- **LTXVideo**:  
   - optimizations: full offload, quantization and tiling support  
   - [TeaCache](https://github.com/ali-vilab/TeaCache/blob/main/TeaCache4LTX-Video/README.md) integration  
-- **VAE**:
+- **VAE**:  
   - tiling granular options in *settings -> variable auto encoder*  
-- **UI**:
+- **UI**:  
   - live preview optimizations and error handling  
   - live preview high quality output, thanks @Disty0  
   - CSS optimizations when log view is disabled  
-- **Samplers**:
+- **Samplers**:  
   - add flow shift options and separate dynamic thresholding from dynamic shifting  
   - autodetect matching sigma capabilities  
-- **API**
+- **API**:  
   - better default values for generate  
-- **Refactor**:
+- **Refactor**:  
   - remove all LDM imports if running in native mode  
-  - startup optimizatios
-- **Fixes**  
+  - startup optimizatios  
+- **Torch**:  
+  - support for `torch==2.6.0`  
+- **Fixes**:  
   - flux pipeline switches: txt/img/inpaint  
   - flux custom unet loader for bnb  
   - interrogate caption with T5  

modules/masking.py

@@ -388,7 +388,7 @@ def run_mask(input_image: Image.Image, input_mask: Image.Image = None, return_ty
     if input_image is None:
         return input_mask
 
-    t0 = time.time()
+    # t0 = time.time()
     input_mask = get_mask(input_image, input_mask) # perform optional auto-masking
     if input_mask is None:
         return None
@@ -436,14 +436,14 @@ def run_mask(input_image: Image.Image, input_mask: Image.Image = None, return_ty
     if opts.invert:
         mask = np.invert(mask)
 
-    mask_size = np.count_nonzero(mask)
-    total_size = np.prod(mask.shape)
-    area_size = np.count_nonzero(mask)
-    t1 = time.time()
 
     return_type = return_type or opts.preview_type
 
-    shared.log.debug(f'Mask: size={input_image.width}x{input_image.height} masked={mask_size}px area={area_size/total_size:.2f} auto={opts.auto_mask} blur={opts.mask_blur:.3f} erode={opts.mask_erode:.3f} dilate={opts.mask_dilate:.3f} type={return_type} time={t1-t0:.2f}')
+    # mask_size = np.count_nonzero(mask)
+    # total_size = np.prod(mask.shape)
+    # area_size = np.count_nonzero(mask)
+    # t1 = time.time()
+    # shared.log.debug(f'Mask: size={input_image.width}x{input_image.height} masked={mask_size}px area={area_size/total_size:.2f} auto={opts.auto_mask} blur={opts.mask_blur:.3f} erode={opts.mask_erode:.3f} dilate={opts.mask_dilate:.3f} type={return_type} time={t1-t0:.2f}')
     if return_type == 'None':
         return input_mask
     elif return_type == 'Opaque':