QLoRA_french_sft/train_results.json

{
    "epoch": 3.0,
    "total_flos": 8.660062333952852e+18,
    "train_loss": 0.0,
    "train_runtime": 0.012,
    "train_samples_per_second": 1934774.909,
    "train_steps_per_second": 1934774.909
}