first_layer_2/model_15000/training_state.json

{
    "global_step": 15000,
    "update_step": 15000,
    "tokens_seen": 9584689152,
    "tokens_seen_before": 9584050176,
    "update_time": 2.877997398376465,
    "wandb_id": "f3ljzhyw"
}