{ "epoch": 7.917525773195877, "eval_audio_cosine_sim": 0.5995970368385315, "eval_loss": 3.09893536567688, "eval_runtime": 1924.8031, "eval_samples": 12, "eval_samples_per_second": 0.006, "eval_steps_per_second": 0.006, "eval_text_cosine_sim": 0.39815211296081543, "total_flos": 962957869218000.0, "train_loss": 8.715420136849085, "train_runtime": 7523.7142, "train_samples": 97, "train_samples_per_second": 0.103, "train_steps_per_second": 0.006 }