| { | |
| "zero_optimization": { | |
| "stage": 2, | |
| "allgather_partitions": true, | |
| "reduce_scatter": true, | |
| "overlap_comm": false, | |
| "contiguous_gradients": true, | |
| "reduce_bucket_size": 5e8, | |
| "allgather_bucket_size": 5e8 | |
| }, | |
| "bf16": {"enabled": "auto"}, | |
| "gradient_clipping": 1.0, | |
| "train_micro_batch_size_per_gpu": "auto", | |
| "gradient_accumulation_steps": "auto", | |
| "train_batch_size": "auto", | |
| "steps_per_print": 50, | |
| "wall_clock_breakdown": false | |
| } |