solved gpu to cpu teething issues

colpali_manager.py

@@ -17,7 +17,6 @@ import spaces
 
 
 #this part is for local runs
-torch.cpu.empty_cache()
 
 model_name = "vidore/colSmol-256M"
 device = get_torch_device("cpu") #try using cpu instead of cpu?

rag.py

@@ -71,9 +71,7 @@ class Rag:
 
         #ollama method below
         
-        torch.cuda.empty_cache() #release cuda so that ollama can use gpu!
-
-    
+           
         os.environ['OLLAMA_FLASH_ATTENTION'] = '1'
         
 
@@ -144,4 +142,4 @@ class Rag:
 #     query = "Based on attached images, how many new cases were reported during second wave peak"
 #     imagesPaths = ["covid_slides_page_8.png", "covid_slides_page_8.png"]
     
-#     rag.get_answer_from_gemini(query, imagesPaths)
+#     rag.get_answer_from_gemini(query, imagesPaths)