#7522 W0921 19:22:36.052000 20400 Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects

2408**442b Posted at: 1 hour ago

W0921 19:22:36.052000 20400 Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.
Traceback (most recent call last):
File "E:\AI\index-tts-win-20250718\webui.py", line 46, in

tts = IndexTTS(model_dir=cmd_args.model_dir, cfg_path=os.path.join(cmd_args.model_dir, "config.yaml"),)

File "E:\AI\index-tts-win-20250718\runtime..\indextts\infer.py", line 79, in init

self.gpt.eval().half()

File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 1180, in half

return self._apply(lambda t: t.half() if t.is_floating_point() else t)

File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 915, in _apply

module._apply(fn)

File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 915, in _apply

module._apply(fn)

File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 915, in _apply

module._apply(fn)

[Previous line repeated 1 more time]
File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 942, in _apply

param_applied = fn(param)

File "E:\AI\index-tts-win-20250718\runtime./Lib/site-packages\torch\nn\modules\module.py", line 1180, in

return self._apply(lambda t: t.half() if t.is_floating_point() else t)

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Press any key to continue . . .
显卡为RTX5090 image

1 Replies

#1 DeepSeek reply at 1 hour ago

核心原因是 Index-TTS 独立部署包自带的 PyTorch 版本过旧,其 CUDA 内核不包含 RTX 50 系(Blackwell 架构,sm_120)的编译镜像,因此 self.gpt.eval().half() 在初始化时直接触发 no kernel image is available 崩溃。这与 pyVideoTrans 本体无关,是 Index-TTS 运行环境本身的 CUDA 兼容问题。

解决方向(按优先级):

  1. 升级 Index-TTS 目录内的 PyTorch(推荐)。RTX 5090 需要 PyTorch 2.7+(CUDA 12.8 及以上) 才包含 sm_120 内核。在 E:\AI\index-tts-win-20250718\runtime 环境下执行:

    runtime\python.exe -m pip install --upgrade torch torchaudio --index-url https://download.pytorch.org/whl/cu128

    升级后重新启动 webui.py

  2. 强制使用 CPU 运行(临时验证)。若是预打包整合包且无法自行升级 PyTorch,可在 webui.pyinfer.py 中把 .half() 及模型加载设备改为 cpu / float32,牺牲速度换取可用,或改用该项目官方提供的新版本整合包(新版本镜像源通常已适配 50 系)。
  3. 改换支持 50 系的 TTS 渠道。若不愿折腾独立部署,可改用 pyVideoTrans 主界面配音渠道中已内置且适配较新 PyTorch 的渠道,如 Qwen3-TTS(内置)F5-TTS(内置)OmniVoice-TTS(内置)Confucius-TTS(内置) 等,避免独立环境问题。
    详见文档:https://pyvideotrans.com/peiyinqudao
补充:知识库未收录 RTX 50 系与 Index-TTS 独立部署包的专门适配说明,以上为基于 CUDA 内核报错的通用判断;若升级 PyTorch 后仍报错,请提供升级后的完整报错日志。

请查阅相关文档:

Thinking...
Post Your Reply
If you have more than 50 messages, you will no longer receive an automatic reply. Please ask a new question!
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Dondate via ko-fi.com

Related resource