first_layer_1/model_2000/training_state.json

{
    "global_step": 2000,
    "update_step": 2000,
    "tokens_seen": 1278001152,
    "tokens_seen_before": 1277362176,
    "update_time": 0.9753463268280029,
    "wandb_id": "krzb2185"
}