| { |
| "checkpoint": { |
| "load_universal": false |
| }, |
| "train_batch_size": "auto", |
| "train_micro_batch_size_per_gpu": "auto", |
| "gradient_accumulation_steps": "auto", |
| "gradient_clipping": "auto", |
| "zero_allow_untested_optimizer": true, |
| "bf16": { |
| "enabled": "auto" |
| }, |
| "communication_data_type": "bf16", |
| "zero_optimization": { |
| "stage": 3, |
| "overlap_comm": true, |
| "contiguous_gradients": true, |
| "sub_group_size": 1e9, |
| "reduce_bucket_size": 5e8, |
| "allgather_bucket_size": 5e8, |
| "stage3_prefetch_bucket_size": 5e8, |
| "stage3_param_persistence_threshold": 1e6, |
| "stage3_max_live_parameters": 1e9, |
| "stage3_max_reuse_distance": 1e9, |
| "stage3_gather_16bit_weights_on_model_save": true |
| }, |
| "activation_checkpointing": { |
| "partition_activations": true, |
| "cpu_checkpointing": true, |
| "number_of_training_steps": 200000, |
| "synchronize_checkpoint_boundary": false, |
| "profile": false |
| } |
| } |
|
|
