MPS_NOT_AVAILABLE pytorch config_error ai_generated true

运行时错误：MPS后端不可用。请安装支持MPS的PyTorch或使用CPU/CUDA。

RuntimeError: MPS backend is not available. Please install PyTorch with MPS support or use CPU/CUDA.

ID: pytorch/mps-backend-unavailable

其他格式: JSON · Markdown 中文 · English
85%修复率
90%置信度
1证据数
2023-04-05首次发现

版本兼容性

版本状态引入弃用备注
torch>=1.12 active — — —
macOS>=12.3 active — — —
MPS>=1.0 active — — —

根因分析

PyTorch编译时未启用MPS（Metal Performance Shaders）支持，或系统缺少兼容的Apple GPU/macOS版本。

English

PyTorch was built without MPS (Metal Performance Shaders) support, or the system lacks a compatible Apple GPU/macOS version.

generic

官方文档

https://pytorch.org/docs/stable/mps.html

解决方案

  1. 安装支持MPS的PyTorch：pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/mps。然后验证：import torch; print(torch.backends.mps.is_available())
  2. 回退到CPU：device = torch.device('cuda' if torch.cuda.is_available() else ('mps' if torch.backends.mps.is_available() else 'cpu'))

无效尝试

常见但无效的做法:

  1. conda install pytorch -c pytorch 70% 失败

    Installing PyTorch from conda default channel often ships CPU-only build; user must explicitly select MPS variant.

  2. export PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 90% 失败

    Setting environment variable PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 does not enable MPS; it only controls memory threshold.