实时推理报错

input device:1:麦克风 (Meizu HiFi DAC Headphone A (MME)
output device:5:耳机 (Meizu HiFi DAC Headphone Am (MME)
{'hubert_path': 'C:/AIGC/TTS/RVC-beta/hubert_base.pt', 'pth_path': 'C:/Users/HP/Documents/RVC_paimon/paimeng/paimeng.pth', 'index_path': 'C:/Users/HP/Documents/RVC_paimon/paimeng/trained_IVF13060_Flat_nprobe_1.index', 'npy_path': 'C:/Users/HP/Documents/RVC_paimon/paimeng/total_fea.npy', 'pitch': 12.0, 'samplerate': 44100, 'block_time': 1.0, 'buffer_num': 1, 'threhold': -30.0, 'crossfade_time': 0.08, 'extra_time': 0.05, 'I_noise_reduce': False, 'O_noise_reduce': False, 'index_rate': 0.5}
using_cuda:True
index search enabled
load model(s) from C:/AIGC/TTS/RVC-beta/hubert_base.pt
gin_channels: 256 self.spk_embed_dim: 109

input_wav:(50362,)
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
请问如何解决实时推理报错的问题

The report includes device settings and a real-time inference log, but no traceback or source entry point. Reproduce the failure with the shown configuration and capture the complete error output first. Done means identifying a reproducible cause and verifying the real-time inference path works without the reported error.