checkpoints-llama32-1b-mathpros10sfeos_diff_intra_shift-2k_2node / 4000 /optimizer /optimizer_config.json
| {"type": "OptimizerFromGradientAccumulator", "parallelism": {"tp_size": "1", "dp_size": "16", "pp_size": "1", "expert_parallel_size": "1"}, "configs": {}} |