Improve error handling while loading GPTQ models

modules/GPTQ_loader.py

@@ -140,7 +140,7 @@ def load_quantized(model_name):
     if shared.args.model_type is None:
         logging.error("The model could not be loaded because its type could not be inferred from its name.")
         logging.error("Please specify the type manually using the --model_type argument.")
-        return
+        return None
 
     # Select the appropriate load_quant function
     model_type = shared.args.model_type.lower()

modules/models.py

@@ -97,7 +97,10 @@ def load_model(model_name):
         model, tokenizer = output
     else:
         model = output
-        tokenizer = load_tokenizer(model_name, model)
+        if model is None:
+            return None, None
+        else:
+            tokenizer = load_tokenizer(model_name, model)
 
     # Hijack attention with xformers
     if any((shared.args.xformers, shared.args.sdp_attention)):