{ "params": 64994816, "tokens": 79999533056, "steps": 152587, "seq_len": 2048, "global_batch_tokens": 524288, "peak_lr": 0.002, "schedule": "wsd-1sqrt", "mix": { "fineweb": 0.45, "dclm": 0.3, "stack": 0.175, "finemath": 0.075 } }