{ "mel_spectrogram": { "n_mel_channels": 100 }, "sampling": { "nfe_step": 32, "cfg_strength": 2, "sway_sampling_coef": -1 }, "backbone": { "depth": 24, "dim": 1024, "heads": 16, "text_dim": 4732 }, "odeint": { "method": "euler" } }