Update generation.py

research/long-context-llama/H2O/generation.py

@@ -59,7 +59,7 @@ if __name__ == '__main__':
         config.hh_size = args.num_heavy_hitter_tokens
         config.recent_size = args.num_local_windows
         config.enable_position_rolling = args.enable_position_rolling
-        model = H2OLlamaForCausalLM.from_pretrained(config)
+        model = H2OLlamaForCausalLM.from_pretrained(model_name, config=config)
     else:
         model = AutoModelForCausalLM.from_pretrained(model_name)