post-training eval_results.json

{
    "eval_samples": 2820,
    "test_accuracy": 0.5429078014184398,
    "test_loss": 1.4821960926055908,
    "test_macro_f1": 0.34723773230887656,
    "test_micro_f1": 0.5429078014184398,
    "test_runtime": 35.207,
    "test_samples_per_second": 80.098,
    "test_steps_per_second": 0.653
}