optimizer/optimizer_config.json

{"type": "OptimizerFromGradientAccumulator", "parallelism": {"tp_size": "1", "dp_size": "256", "pp_size": "1", "expert_parallel_size": "1"}, "configs": {}}