{ "epoch": 2.9690721649484537, "eval_audio_cosine_sim": 0.5888153910636902, "eval_loss": 3.0872552394866943, "eval_runtime": 496.7716, "eval_samples": 12, "eval_samples_per_second": 0.024, "eval_steps_per_second": 0.024, "eval_text_cosine_sim": 0.3267640173435211, "total_flos": 293443795561212.0, "train_loss": 10.172218322753906, "train_runtime": 1046.7105, "train_samples": 97, "train_samples_per_second": 0.278, "train_steps_per_second": 0.017 }