Configuration of graph parallel option is incorrect.
Tensor parallel size or data parallel size is incorrect.
Make sure the options in graph_parallel_option.json are correct:
1. If both tensor_parallel_option and data_parallel_option are set, the total parallel size (i.e. tensor_parallel_size * data_parallel_size) must equal the total device num in numa_config.json;
2. If only either tensor_parallel_option or data_parallel_option is set, the tensor_parallel_size or data_parallel_option must equal the total device num in numa_config,json.