llama-3.2-3b-sft/train_results.json

{
    "epoch": 1.0,
    "total_flos": 2.1312002805936947e+18,
    "train_loss": 1.184760999939929,
    "train_runtime": 31570.962,
    "train_samples": 117772,
    "train_samples_per_second": 1.949,
    "train_steps_per_second": 0.122
}