{ "generated_vs_real": 67.00966902642759, "griffinlim_floor_vs_real": 60.52624444235573, "no_prompt_vs_real": 87.13893072220702, "config": { "n_eval": 300, "cfg": 4.0, "steps": 50, "pann_sample_rate": 32000, "note": "upstream frechet_audio_distance PANN path has a shape bug (np.concatenate instead of stack on 1D per-clip embeddings); worked around here by reshaping to (-1, 2048) before stats" } }