Small fixes

lm_eval/models/gpt2.py

@@ -15,12 +15,12 @@ class HFLM(BaseLM):
         if device:            
             if device not in ["cuda", "cpu"]:
                 device = int(device)
-            self.device = torch.device(device)
+            self._device = torch.device(device)
             print(f"Using device '{device}'")
         else:
             print("Device not specificed")
             print(f"Cuda Available? {torch.cuda.is_available()}")
-            self.device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
+            self._device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
 
         # TODO: update this to be less of a hack once subfolder is fixed in HF
         self.gpt2 = transformers.AutoModelForCausalLM.from_pretrained(

main.py

@@ -43,6 +43,7 @@ def parse_args():
     parser.add_argument('--decontaminate', action="store_true")
     parser.add_argument('--ngrams_path', default=None)
     parser.add_argument('--ngrams_n_size', type=int, default=None)
+    parser.add_argument('--description_dict_path', default=None)    
 
     return parser.parse_args()
 
@@ -67,11 +68,6 @@ def pattern_match(patterns, source_list):
             task_names.add(matching)
     return list(task_names)
 
-def main():
-    parser.add_argument('--description_dict_path', default=None)
-    return parser.parse_args()
-
-
 def main():
     args = parse_args()
     if not ensure_correct_decontamination_params(args):