Update the usage names in the handler.

handler.py

@@ -63,21 +63,24 @@ class ImagePreprocessor():
 usage_to_weights_file = {
     'General': 'BiRefNet',
     'General-HR': 'BiRefNet_HR',
-    'General-Lite': 'BiRefNet_lite',
-    'General-Lite-2K': 'BiRefNet_lite-2K',
-    'General-reso_512': 'BiRefNet-reso_512',
+    'Matting-HR': 'BiRefNet_HR-matting',
     'Matting': 'BiRefNet-matting',
-    'Matting-HR': 'BiRefNet_HR-Matting',
     'Portrait': 'BiRefNet-portrait',
+    'General-reso_512': 'BiRefNet_512x512',
+    'General-Lite': 'BiRefNet_lite',
+    'General-Lite-2K': 'BiRefNet_lite-2K',
+    # 'Anime-Lite': 'BiRefNet_lite-Anime',
     'DIS': 'BiRefNet-DIS5K',
     'HRSOD': 'BiRefNet-HRSOD',
     'COD': 'BiRefNet-COD',
     'DIS-TR_TEs': 'BiRefNet-DIS5K-TR_TEs',
-    'General-legacy': 'BiRefNet-legacy'
+    'General-legacy': 'BiRefNet-legacy',
+    'General-dynamic': 'BiRefNet_dynamic',
+    'Matting-dynamic': 'BiRefNet_dynamic-matting',
 }
 
 # Choose the version of BiRefNet here.
-usage = 'General'
+usage = 'General-dynamic'
 
 # Set resolution
 if usage in ['General-Lite-2K']:
@@ -87,7 +90,11 @@ elif usage in ['General-reso_512']:
 elif usage in ['General-HR', 'Matting-HR']:
     resolution = (2048, 2048)
 else:
-    resolution = (1024, 1024) 
+    if '-dynamic' in usage:
+        resolution = None
+        print('Using the original size (div by 32) for inference.')
+    else:
+        resolution = (1024, 1024)
 
 half_precision = True
 
@@ -123,6 +130,10 @@ class EndpointHandler():
 
         image = image_ori.convert('RGB')
         # Preprocess the image
+        if resolution is None:
+            resolution_div_by_32 = [int(int(reso)//32*32) for reso in image.size]
+            if resolution_div_by_32 != resolution:
+                resolution = resolution_div_by_32
         image_preprocessor = ImagePreprocessor(resolution=tuple(resolution))
         image_proc = image_preprocessor.proc(image)
         image_proc = image_proc.unsqueeze(0)

requirements.txt

@@ -1,4 +1,4 @@
-torch==2.5.1
+torch>=2.5.1
 torchvision
 numpy<2
 opencv-python