RuntimeError: expected scalar type ComplexDouble but found ComplexHalf[BUG:API]

[panpan123456]

阅读 README.md 和 dependencies.md

• 我已经阅读过 README.md 和 dependencies.md 文件

检索 issue 和 discussion

• 我已经确认之前没有 issue 或 discussion 涉及此 BUG

检查 Forge 版本

• 我已经确认问题发生在最新代码或稳定版本中

Forge Commit 或者 Tag

1

Python 版本

3.11

PyTorch 版本

2.3.1

操作系统信息

macox

BUG 描述

    with torch.no_grad():

        if not skip_refine_text:
            refined = self.instance._refine_text(
                text,
                self.instance.device,
                params_refine_text,
            )
            text_tokens = refined.ids
            text_tokens = [
                i[i.less(self.instance.tokenizer_break_0_ids)] for i in text_tokens
            ]
            text = self.get_tokenizer().batch_decode(text_tokens)
            refined.destroy()
            if refine_text_only:
                yield text
                return
        if not smooth_decoding:
            length = [0 for _ in range(len(text))]
            for result in self.instance._infer_code(
                text,
                stream,
                self.instance.device,
                use_decoder,
                params_infer_code,
            ):
                wavs = self._decode_to_wavs(result, length, use_decoder)
                yield wavs

代码运行到这里会报错：
RuntimeError: expected scalar type ComplexDouble but found ComplexHalf

BUG 端点

tts_to_audio/tts_stream

复现参数

{
"text": "这是一个很有意思的事情",
"speaker_wav": "3d40607bf216486fb508758fa67f1a83",
"language": "zh"
}

期望结果

返回音频

实际结果

报错

错误信息

File "/Users/panhong/miniconda3/envs/ChatTTS_colab/lib/python3.11/site-packages/vocos/spectral_ops.py", line 46, in forward
    return torch.istft(spec, self.n_fft, self.hop_length, self.win_length, self.window, center=True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: expected scalar type ComplexDouble but found ComplexHalf

[zhzLuke96]

试试启动时带上 --no_half 参数吧，目前mps的适配不是很好，可能没法支持半精度（默认半精度）

[panpan123456]

done