Update main.py

main.py

@@ -7,7 +7,9 @@ from transformers import AutoTokenizer, AutoModel
 import torch
 from torch import Tensor
 import torch.nn.functional as F
-
+from Allam_Backend_HF import (
+    allam_llm
+    ) 
 import os
 os.environ['HF_HOME'] = '/'
 
@@ -56,9 +58,7 @@ def e5_embeddings(query: str = Query(...)):
     else:
         raise HTTPException(status_code=500)
 
-from Allam_Backend import (
-    allam_llm
-    ) 
+
 @app.get("/allam_response")
 def allam_response(query: str = Query(...)):