eval_results.json

{
    "epoch": 2.9690721649484537,
    "eval_audio_cosine_sim": 0.5567758679389954,
    "eval_loss": 3.091003179550171,
    "eval_runtime": 1153.212,
    "eval_samples": 12,
    "eval_samples_per_second": 0.01,
    "eval_steps_per_second": 0.01,
    "eval_text_cosine_sim": 0.46733546257019043
}