update CLIP loda device

llava/model/multimodal_encoder/clip_encoder.py

@@ -27,7 +27,7 @@ class CLIPVisionTower(nn.Module):
             return
 
         self.image_processor = CLIPImageProcessor.from_pretrained(self.vision_tower_name)
-        self.vision_tower = CLIPVisionModel.from_pretrained(self.vision_tower_name, device_map=device_map)
+        self.vision_tower = CLIPVisionModel.from_pretrained(self.vision_tower_name, device_map='cuda')# Error on the HuggingFace for auto
         self.vision_tower.requires_grad_(False)
 
         self.is_loaded = True