{ "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, "total_flos": 284409215227248.0, "train_loss": 9.754847163245792, "train_runtime": 2969.0391, "train_samples": 125, "train_samples_per_second": 0.126, "train_steps_per_second": 0.007 }