daft-qwen2.5-coder-3b-instruct-full-loss-0.02/all_results.json

{
    "epoch": 3.0,
    "total_flos": 1577088536150016.0,
    "train_loss": 0.06165807318099385,
    "train_runtime": 23128.4215,
    "train_samples_per_second": 2.881,
    "train_steps_per_second": 0.045
}