Added VRAM and RAM cleanup after request is finished.

svc/router.py

@@ -10,6 +10,7 @@ import os
 import json
 from src.deepeval.deepeval_task_manager import DeepEvalTaskManager
 import torch
+import gc
 from time import time
 from huggingface_hub import HfApi, ModelInfo
 
@@ -111,9 +112,15 @@ async def deep_eval_suite(request: DeepEvalSuiteRequest):
         "end_time": end_time
     }
 
-
     json_results = json.dumps(tbr_dict)
 
+    #Free up VRAM
+    torch.cuda.empty_cache()
+
+    #Free up RAM
+    des = None
+    gc.collect()
+
     return TaskResponse(results=json_results)