diff --git a/ds_config_zero3_lora.json b/ds_config_zero3_lora.json index 0e53a51..940531e 100644 --- a/ds_config_zero3_lora.json +++ b/ds_config_zero3_lora.json @@ -9,17 +9,13 @@ "stage": 3, "overlap_comm": true, "contiguous_gradients": true, - "allgather_partitions": true, - "reduce_scatter": true, - "round_robin_gradients": true, - "reduce_bucket_size": 150000000, - "stage3_prefetch_bucket_size": 100000000, + "reduce_bucket_size": 500000000, + "stage3_prefetch_bucket_size": 200000000, "stage3_param_persistence_threshold": 1000000, "offload_optimizer": { "device": "none" }, - "offload_param": { "device": "none" } + "offload_param": { "device": "none" } }, - - "stage3_gather_16bit_weights_on_model_save": false, + "gradient_clipping": 1.0, "wall_clock_breakdown": false }