attempt to revert to embedding fix 2

app/services/embedding_service.py

@@ -25,8 +25,7 @@ class EmbeddingService:
         self.tokenizer = AutoTokenizer.from_pretrained(model_name)
         self.model = AutoModel.from_pretrained(model_name)
         
-        # Move model to CPU and eval mode
-        self.model = self.model.to("cpu")
+        
         self.model.eval()
         
         # Get embedding dimension from model config