train_results.json

{
 "epoch": 3.0,
 "num_input_tokens_seen": 2965968,
 "total_flos": 1.2618658999382835e+17,
 "train_loss": 0.45766232585906985,
 "train_runtime": 3803.7297,
 "train_samples_per_second": 1.577,
 "train_steps_per_second": 0.099
}