Kokoro-82M can't use GPU？ #4029
Description
System Info / 系統信息
system: ubuntu22.04
CUDA Version: 12.8
python: 3.11
Running Xinference with Docker? / 是否使用 Docker 运行 Xinfernece？
- docker / docker
- pip install / 通过 pip install 安装
- installation from source / 从源码安装
Version info / 版本信息
1.9.1
The command used to start Xinference / 用以启动 xinference 的命令
xinference launch --model-name Kokoro-82M --model-type audio --replica 1 --n-gpu auto --gpu-idx 0 --download-hub modelscope --lang_code z
Reproduction / 复现过程
(base) javis@javis-MS-7D91:~/Project/xinference$ nvidia-smi
Mon Sep 1 14:33:28 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.133.07 Driver Version: 570.133.07 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A6000 Off | 00000000:01:00.0 Off | Off |
| 30% 46C P2 68W / 300W | 37740MiB / 49140MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 1326 G /usr/lib/xorg/Xorg 19MiB |
| 0 N/A N/A 2369 C GPT_SoVITS 2056MiB |
| 0 N/A N/A 2087717 C ...lication/anaconda3/bin/python 19390MiB |
| 0 N/A N/A 2091269 C ...lication/anaconda3/bin/python 12562MiB |
| 0 N/A N/A 2567921 C Model: bge-m3-0 2446MiB |
| 0 N/A N/A 2568268 C Model: SenseVoiceSmall-0 1226MiB |
+-----------------------------------------------------------------------------------------+
GPU未加载Kokoro-82M
Expected behavior / 期待表现
修改xinference/model/audio/kokoro.p 第84行代码后正常
82 self._model = KPipeline(
83 lang_code=lang_code,
84 model=KModel(config=config_path, model=model_path).to(self._device),
85 device=self._device,
86 )Mon Sep 1 14:51:07 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.133.07 Driver Version: 570.133.07 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A6000 Off | 00000000:01:00.0 Off | Off |
| 33% 63C P2 72W / 300W | 38385MiB / 49140MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 1326 G /usr/lib/xorg/Xorg 19MiB |
| 0 N/A N/A 2369 C GPT_SoVITS 2056MiB |
| 0 N/A N/A 2087717 C ...lication/anaconda3/bin/python 19390MiB |
| 0 N/A N/A 2091269 C ...lication/anaconda3/bin/python 12562MiB |
| 0 N/A N/A 2567921 C Model: bge-m3-0 2446MiB |
| 0 N/A N/A 2568268 C Model: SenseVoiceSmall-0 1226MiB |
| 0 N/A N/A 2611921 C Model: Kokoro-82M-0 640MiB |
+-----------------------------------------------------------------------------------------+