eval_results.json

{
    "epoch": 2.9690721649484537,
    "eval_audio_cosine_sim": 0.5600753426551819,
    "eval_loss": 3.0910282135009766,
    "eval_runtime": 1162.6673,
    "eval_samples": 12,
    "eval_samples_per_second": 0.01,
    "eval_steps_per_second": 0.01,
    "eval_text_cosine_sim": 0.4803898334503174
}