Update gen.py

gen.py

@@ -67,7 +67,7 @@ def get_pretrained_models(
     with open(Path(llama_weight_path) / "params.json", "r") as f:
         params = json.loads(f.read())
 
-    model_args: ModelArgs = ModelArgs(max_seq_len=1024, max_batch_size=1, **params)
+    model_args: ModelArgs = ModelArgs(max_seq_len=512, max_batch_size=1, **params)
     tokenizer = Tokenizer(model_path=f"{tokenizer_weight_path}/tokenizer.model")
     model_args.vocab_size = tokenizer.n_words
     torch.set_default_tensor_type(torch.cuda.HalfTensor)