无法打开TTS推理的WebUI且控制台报错：RuntimeError: CUDA error: no kernel image is available for execution on the device

[MrA380]

使用WebUI的1-GPT-SoVITS_TTS中的1c-推理模块时： 无法正常打开TTS推理的WebUI 且控制台（日志）报错：RuntimeError: CUDA error: no kernel image is available for execution on the device 环境：Windows 10 22H2\已配置ffmpeg、ffprobe\GPU：(N)GeForce GTX 650

[MrA380]

• Console

"F:\ai-nina\GPT-SoVITS-v2-240821\runtime\python.exe" GPT_SoVITS/inference_webui_fast.py "zh_CN" ---------------------------------------------TTS Config--------------------------------------------- device : cuda is_half : True version : v2 t2s_weights_path : GPT_weights/芹井仁菜V3-e15.ckpt vits_weights_path : SoVITS_weights/芹井仁菜V3_e100_s1800.pth bert_base_path : GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large cnhuhbert_base_path : GPT_SoVITS/pretrained_models/chinese-hubert-base

Loading Text2Semantic weights from GPT_weights/芹井仁菜V3-e15.ckpt Traceback (most recent call last): File "F:\ai-nina\GPT-SoVITS-v2-240821\GPT_SoVITS\inference_webui_fast.py", line 106, in tts_pipeline = TTS(tts_config) File "F:\ai-nina\GPT-SoVITS-v2-240821\GPT_SoVITS\TTS_infer_pack\TTS.py", line 243, in init self._init_models() File "F:\ai-nina\GPT-SoVITS-v2-240821\GPT_SoVITS\TTS_infer_pack\TTS.py", line 268, in _init_models self.init_t2s_weights(self.configs.t2s_weights_path) File "F:\ai-nina\GPT-SoVITS-v2-240821\GPT_SoVITS\TTS_infer_pack\TTS.py", line 346, in init_t2s_weights self.t2s_model = self.t2s_model.half() File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\lightning_fabric\utilities\device_dtype_mixin.py", line 99, in half return super().half() File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\torch\nn\modules\module.py", line 1001, in half return self._apply(lambda t: t.half() if t.is_floating_point() else t) File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\torch\nn\modules\module.py", line 797, in _apply module._apply(fn) File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\torch\nn\modules\module.py", line 797, in _apply module._apply(fn) File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\torch\nn\modules\module.py", line 820, in _apply param_applied = fn(param) File "F:\ai-nina\GPT-SoVITS-v2-240821\runtime\lib\site-packages\torch\nn\modules\module.py", line 1001, 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.

[SapphireLab]

原因可能是整合包中的 torch 使用的是对应 CUDA 11.8 的版本, CUDA 11.8 支持的显卡算力应在 3.5 以上, 而 GTX 650 只有 3.0. 可参阅 https://en.wikipedia.org/wiki/CUDA#GPUs_supported.

• 可以考虑用 CPU 运行 (简单的修改如下):
  • 在 ~\GPT-SoVITS-v2-240821\GPT_SoVITS\inference_webui.py 中 100 行左右，输入 device="cpu"；
  • 类似地，勾选并行推理时，在 ~\GPT-SoVITS-v2-240821\GPT_SoVITS\inference_webui_fast.py 中 60 行左右，输入 device="cpu" 即可.
• 或者自行降级环境以兼容 GPU, 但版本比较过时不能保证全部顺利运行.
• 换个设备.

[MrA380]

感谢！