Update models/model_loader.py

models/model_loader.py

@@ -5,6 +5,7 @@ CACHE_DIR = "/tmp/pretrained_models"
 
 import torch
 import gc
+import time
 
 from speechbrain.pretrained.interfaces import foreign_class
 from faster_whisper import WhisperModel
@@ -41,3 +42,4 @@ def unload_model(model):
     del model
     torch.cuda.empty_cache()
     gc.collect()
+    time.sleep(10)  # give system time to clean up before moving on