{ "final_step": 274555, "supervised_tokens_seen": 26983265400, "actual_target_supervised_tokens": 26983343205, "target_supervised_tokens_requested": 27000000000, "params": 75054720, "label_policy": "labels=input_ids; LlamaForCausalLM shifts internally", "data_order": "sequential contiguous token blocks", "block_size": 4096, "supervised_targets_per_sequence": 4095, "batch_size": 24, "grad_accum": 1, "param_dtype": "bf16", "compute_dtype": "bf16 autocast" }