train_results.json

{
    "epoch": 2.96,
    "num_input_tokens_seen": 252064,
    "total_flos": 1.2630237909123072e+16,
    "train_loss": 0.40682300817859063,
    "train_runtime": 7468.1424,
    "train_samples_per_second": 0.121,
    "train_steps_per_second": 0.015
}