Server 500 when using the TTS endpoint

[rampadc]

Using version 0.3.2, installed on a fresh Python 3.12 environment. I ran the audio.ipynb notebook and got server 500 errors on every request.

Python output:

InternalServerError: Error code: 500 - {'detail': 'Error reading audio file: arange(): incompatible function arguments. The following argument types are supported:\n    1. arange(start : Union[int, float], stop : Union[int, float], step : Union[None, int, float], dtype: Optional[Dtype] = None, *, stream: Union[None, Stream, Device] = None) -> array\n    2. arange(stop : Union[int, float], step : Union[None, int, float] = None, dtype: Optional[Dtype] = None, *, stream: Union[None, Stream, Device] = None) -> array\n\nInvoked with types: mlx.core.array'}

mlx-omni-server outputs:

Got duration of 757 frames (8.150339126586914 secs) for generated speech.
INFO:     127.0.0.1:62402 - "POST /v1/audio/speech HTTP/1.1" 500 Internal Server Error
Fetching 5 files: 100%|████████████████████████| 5/5 [00:00<00:00, 49461.13it/s]
Fetching 2 files: 100%|████████████████████████| 2/2 [00:00<00:00, 22733.36it/s]