post-training eval_results.json

{
    "eval_samples": 2737,
    "test_accuracy": 0.7168432590427475,
    "test_loss": 0.8912712335586548,
    "test_macro_f1": 0.5103889782929176,
    "test_micro_f1": 0.7168432590427475,
    "test_runtime": 5.8747,
    "test_samples_per_second": 465.9,
    "test_steps_per_second": 3.745
}