eval_results.json

{
    "epoch": 2.6879999999999997,
    "eval_audio_cosine_sim": 0.6386334300041199,
    "eval_loss": 3.099665880203247,
    "eval_runtime": 1718.2675,
    "eval_samples": 12,
    "eval_samples_per_second": 0.007,
    "eval_steps_per_second": 0.007,
    "eval_text_cosine_sim": 0.42658713459968567
}