| compute_environment: LOCAL_MACHINE | |
| distributed_type: DEEPSPEED | |
| deepspeed_config: | |
| deepspeed_config_file: configs/ds_config.json | |
| zero3_init_flag: true | |
| deepspeed_hostfile: /mnt/jfzn/msj/flash-linear-attention/legacy/training/hostfile.txt | |
| deepspeed_multinode_launcher: pdsh | |
| machine_rank: 0 | |
| main_process_ip: "10.119.141.222" | |
| main_process_port: 29500 | |
| main_training_function: main | |
| num_machines: 4 | |
| num_processes: 32 | |
| same_network: true | |
| use_cpu: false | |
| rdzv_backend: c10d | |
| tpu_env: [] | |
| tpu_use_cluster: false | |
| tpu_use_sudo: false | |
| # deepspeed_multinode_launcher: slurm | |
| # 10.119.141.212 51003 | |
| # 10.119.141.222 51009 |