{ "epoch": 2.6879999999999997, "eval_audio_cosine_sim": 0.5799151062965393, "eval_loss": 3.0975494384765625, "eval_runtime": 1080.4689, "eval_samples": 12, "eval_samples_per_second": 0.011, "eval_steps_per_second": 0.011, "eval_text_cosine_sim": 0.49306759238243103, "total_flos": 247581948423984.0, "train_loss": 9.755998475211006, "train_runtime": 4057.189, "train_samples": 125, "train_samples_per_second": 0.092, "train_steps_per_second": 0.005 }