fix: import create_model_from_pretrained

models/siglip_model.py

@@ -40,15 +40,15 @@ class SigLIPModel:
                 print(f"Warning: Checkpoint not found at {self.ckpt_path}")
             
             if 'hf' in self.ckpt_path:
-                self.ckpt_path = hf_hub_download("timm/ViT-SO400M-14-SigLIP-384", "open_clip_pytorch_model.bin")
-                # self.tokenizer = open_clip.get_tokenizer("hf-hub:timm/ViT-SO400M-14-SigLIP-384")
-                self.tokenizer = HFTokenizer("hf-hub:timm/ViT-SO400M-14-SigLIP-384")
-            else:
-                self.tokenizer = HFTokenizer(self.tokenizer_path)
-            self.model, _, self.preprocess = open_clip.create_model_and_transforms(
-                self.model_name, 
-                pretrained=self.ckpt_path
-            )
+            from open_clip import create_model_from_pretrained, get_tokenizer # works on open-clip-torch>=2.23.0, timm>=0.9.8
+            
+            model, preprocess = create_model_from_pretrained('hf-hub:timm/ViT-SO400M-14-SigLIP-384')
+            tokenizer = get_tokenizer('hf-hub:timm/ViT-SO400M-14-SigLIP-384')
+            
+            # model, preprocess = create_model_from_pretrained('hf-hub:timm/ViT-SO400M-14-SigLIP-384')
+            # from huggingface_hub import hf_hub_download
+            # tokenizer = get_tokenizer('hf-hub:timm/ViT-SO400M-14-SigLIP-384')
+
             self.model = self.model.to(self.device)
             self.model.eval()