add add assertion to build function of Llama

llama/generation.py

@@ -64,6 +64,10 @@ def build(
  This method initializes the distributed process group, sets the device to CUDA,
  and loads the pre-trained model and tokenizer.
  """
+ assert 1 <= max_seq_len <= 8192, f"max_seq_len must be between 1 and 8192, got {max_seq_len}."
+ assert os.path.isdir(ckpt_dir), f"Checkpoint directory '{ckpt_dir}' does not exist."
+ assert os.path.isfile(tokenizer_path), f"Tokenizer file '{tokenizer_path}' does not exist."
+
  if not torch.distributed.is_initialized():
  torch.distributed.init_process_group("nccl")
  if not model_parallel_is_initialized():