伍肆參

手上 Notebook A 是 , Notebook B 卻是 , 又有無 GPU 的Notebook ...
.....嗯嗯... 啥都有~ 就是沒有 :p
由於常會更換機器使用，需要一個能自動判定/選擇最佳GPU 的
網路上找許久都不見有人提供相似功能
在此紀錄一下，期待日後介面統一的到來

另，期待能人志士添加 AMD GPU判定辦法
暫時使用方式偵測 AMD GPU, 功能尚未確認過
AMD使用 DirectML方式需先安裝：

pip install torch-directml

非AMD用戶不建議服用，極大可能會造成身體不適，氣血失調，驚奇不順斷：

torch-directml 0.2.5.dev240914 requires torch==2.4.1, but you have torch 2.8.0 which is incompatible.
torch-directml 0.2.5.dev240914 requires torchvision==0.19.1, but you have torchvision 0.23.0 which is incompatible.

--
Intel GPU 需要先安裝 Intel Extension for PyTorch (IPEX)

```

python -m pip install torch torchvision torchaudio intel_extension_for_pytorch --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/

```

驗證(after reboot)：
```

python -c "import torch; import intel_extension_for_pytorch as ipex; print(ipex.__version__)"

```

"import intel_extension_for_pytorch as ipex" 時，可能會遇到：
```

ERROR! Intel® Extension for PyTorch* needs to work with PyTorch 2.8.*, but PyTorch 2.9.1+cpu is found. Please switch to the matching version and run again.

``` 錯誤

>

pip install torch==2.8

``` torchaudio 2.9.1 requires torch==2.9.1, but you have torch 2.8.0 which is incompatible.
torchvision 0.24.1 requires torch==2.9.1, but you have torch 2.8.0 which is incompatible.```

>

pip install torch==2.8 torchaudio==2.8 torchvision==0.23.0

或是遇到： ``` OSError: [WinError 126] The specified module could not be found.
Error loading "C:\.....\intel_extension_for_pytorch\bin\esimd_kernels.dll" or one of its dependencies. ```
Ref: https://github.com/intel/intel-extension-for-pytorch/issues/804#issuecomment-2840629788 (IE Only)

python -m pip install torch==2.8.0+xpu torchaudio torchvision --index-url https://download.pytorch.org/whl/xpu
python -m pip install intel-extension-for-pytorch==2.8.10+xpu --index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/

And, Install Latest supported redistributable version by MS , and reboot

* nVidia 安裝 torch 即可：

pip install torch

CODE:


import torch
import sys

HAS_IPEX = False
HAS_DML = False

# 嘗試導入 Intel Extension for PyTorch
# Try to import Intel Extension for PyTorch
try:
    import intel_extension_for_pytorch as ipex
    HAS_IPEX = True
except ImportError:
    HAS_IPEX = False

# 嘗試導入 PyTorch DirectML (用於 AMD GPU 支援)
try:
    import torch_directml
    HAS_DML = True
except ImportError:
    HAS_DML = False

def get_best_device():
    """
    自動檢測並返回最佳可用設備 (Auto-detect and return best available device).
    優先級 Priority:
    1. NVIDIA GPU (CUDA)
    2. Intel GPU (XPU)
    3. AMD GPU (DirectML)
    4. CPU
    """
    
    # 1. 檢查 NVIDIA CUDA
    if torch.cuda.is_available():
        print(f"[設備] 偵測到 NVIDIA GPU: {torch.cuda.get_device_name(0)}")
        return torch.device("cuda")
    
    # 2. 檢查 Intel XPU (Arc / Iris Xe)
    # PyTorch 2.4+ 可能直接支援 torch.xpu，但通常需要 IPEX
    if HAS_IPEX and hasattr(torch, 'xpu') and torch.xpu.is_available():
        print(f"[設備] 偵測到 Intel GPU (XPU): {torch.xpu.get_device_name(0)}")
        return torch.device("xpu")
    
    # 用於舊版 IPEX 的備用方案
    if HAS_IPEX and hasattr(ipex, 'xpu') and ipex.xpu.is_available():
         print(f"[設備] 偵測到 Intel GPU (XPU via IPEX)")
         return torch.device("xpu")
    # 3. 檢查 AMD GPU (DirectML)
    if HAS_DML and torch_directml.is_available():
        # DirectML 設備 ID 通常為 0
        dml_device = torch_directml.device()
        print(f"[設備] 偵測到 AMD/DirectML GPU: {torch_directml.device_name(0)}")
        return dml_device
    # 4. 降級至 CPU
    print("[設備] 未偵測到 GPU。使用 CPU。")
    return torch.device("cpu")

def print_system_info():
    print(f"Python Version: {sys.version}")
    print(f"PyTorch Version: {torch.__version__}")
    print(f"Intel Extension (IPEX) 已安裝: {'是' if HAS_IPEX else '否'}")
    print(f"PyTorch DirectML (AMD) 已安裝: {'是' if HAS_DML else '否'}")
    
    if torch.cuda.is_available():
        print(f"CUDA 可用: 是, 版本: {torch.version.cuda}")
        print(f"GPU 名稱: {torch.cuda.get_device_name(0)}")
    else:
        print("CUDA 可用: 否")
    if HAS_IPEX:
        try:
            # 註: IPEX 版本檢查方式可能有所不同
            print(f"IPEX 版本: {ipex.__version__}")
        except:
            pass
        if hasattr(torch, 'xpu') and torch.xpu.is_available():
             print(f"XPU 可用: 是")
             print(f"XPU 名稱: {torch.xpu.get_device_name(0)}")
             
    if HAS_DML and torch_directml.is_available():
        print(f"DirectML 可用: 是")
        print(f"DirectML 設備: {torch_directml.device_name(0)}")

if __name__ == "__main__":
    print("=== Hardware Detection Info ===")
    print_system_info()
    print("\n=== Selecting Device ===")
    device = get_best_device()
    print(f"Selected Device: {device}")
    # Example Usage in Training
    # model = MyModel().to(device)
    # data = data.to(device)

* :


Python Version: 3.12.3 | packaged by conda-forge | (main, Apr 15 2024, 18:20:11) [MSC v.1938 64 bit (AMD64)]
PyTorch Version: 2.8.0+xpu
Intel Extension (IPEX) 已安裝: 是
PyTorch DirectML (AMD) 已安裝: 否
CUDA 可用: 否
IPEX 版本: 2.8.10+xpu
XPU 可用: 是
XPU 名稱: Intel(R) Arc(TM) A350M Graphic

* nVidia RTX4060 GPU :


=== Hardware Detection Info ===
Python Version: 3.10.16 | packaged by Anaconda, Inc. | (main, Dec 11 2024, 16:19:12) [MSC v.1929 64 bit (AMD64)]
PyTorch Version: 2.1.2+cu121
Intel Extension (IPEX) 已安裝: 否
PyTorch DirectML (AMD) 已安裝: 否
CUDA 可用: 是, 版本: 12.1
GPU 名稱: NVIDIA GeForce RTX 4060 Laptop GPU

=== Selecting Device ===
[設備] 偵測到 NVIDIA GPU: NVIDIA GeForce RTX 4060 Laptop GPU
Selected Device: cuda

伍肆參

2026/01/08

[Python] 自動判斷並選定最優 GPU : Intel GPU(Arc A350m) , nVidia GeForce RTX 4060 or CPU

Translate

2026/01/08

[Python] 自動判斷 並選定 最優 GPU : Intel GPU(Arc A350m) , nVidia GeForce RTX 4060 or CPU

[Python] 自動判斷並選定最優 GPU : Intel GPU(Arc A350m) , nVidia GeForce RTX 4060 or CPU