利用Python进行全面的GPU环境检测与分析

python gpu_info.py

import sys
import os
import subprocess
import torch
from datetime import datetime

def get_gpu_info():
    print("=" * 50)
    print("GPU 诊断报告")
    print("=" * 50)
    print(f"诊断时间: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")

    # 系统信息
    print("系统信息:")
    print(f"Python 版本: {sys.version}")
    print(f"PyTorch 版本: {torch.__version__}")
    print(f"CUDA 版本 (PyTorch): {torch.version.cuda}")
    print(f"cuDNN 版本: {torch.backends.cudnn.version()}\n")

    # CUDA 环境检查
    print("CUDA 环境变量:")
    cuda_vars = {
        'CUDA_HOME': os.environ.get('CUDA_HOME', '未设置'),
        'CUDA_PATH': os.environ.get('CUDA_PATH', '未设置'),
        'CUDA_VISIBLE_DEVICES': os.environ.get('CUDA_VISIBLE_DEVICES', '未设置')
    }
    for var, value in cuda_vars.items():
        print(f"{var}: {value}")
    print()

    # NVIDIA-SMI 输出
    print("NVIDIA-SMI 信息:")
    try:
        encodings = ['gbk', 'utf-8', 'iso-8859-1']
        nvidia_smi = None
        for encoding in encodings:
            try:
                nvidia_smi = subprocess.check_output(["nvidia-smi"]).decode(encoding)
                break
            except UnicodeDecodeError:
                continue
        
        if nvidia_smi:
            print(nvidia_smi)
        else:
            print("无法解码 nvidia-smi 输出")
    except Exception as e:
        print(f"执行 nvidia-smi 失败: {e}\n")

    # PyTorch GPU 信息
    print("\nPyTorch GPU 详细信息:")
    if torch.cuda.is_available():
        print(f"检测到 {torch.cuda.device_count()} 个 GPU 设备")
        
        for i in range(torch.cuda.device_count()):
            props = torch.cuda.get_device_properties(i)
            print(f"\nGPU {i}: {props.name}")
            print(f"├─ 计算能力: {props.major}.{props.minor}")
            print(f"├─ 总显存: {props.total_memory / (1024**2):.1f} MB")
            print(f"├─ 多处理器数量: {props.multi_processor_count}")
            print(f"├─ 最大线程数/块: {props.max_threads_per_multi_processor}")
            
            # 显存使用情况
            try:
                memory_allocated = torch.cuda.memory_allocated(i) / (1024**2)
                memory_reserved = torch.cuda.memory_reserved(i) / (1024**2)
                memory_free = (props.total_memory / (1024**2)) - memory_allocated
                
                print(f"├─ 已分配显存: {memory_allocated:.1f} MB")
                print(f"├─ 已预留显存: {memory_reserved:.1f} MB")
                print(f"└─ 可用显存: {memory_free:.1f} MB")
            except Exception as e:
                print(f"└─ 无法获取显存使用情况: {e}")
    else:
        print("未检测到可用的 GPU 设备")
        print("\n可能的原因:")
        print("1. CUDA 驱动未正确安装")
        print("2. PyTorch 未编译 CUDA 支持")
        print("3. GPU 被其他进程占用")
        print("4. 系统未正确识别 GPU")

def get_usb_controller_info():
    print("\nUSB 控制器信息:")
    try:
        result = subprocess.check_output(["wmic", "path", "Win32_USBController", "get", "name,manufacturer"], 
                                      encoding='gbk')
        controllers = result.strip().split('\n')[1:]  # Skip header
        nvidia_controllers = []
        
        for controller in controllers:
            if controller.strip():
                if "NVIDIA" in controller:
                    nvidia_controllers.append(controller.strip())
                    
        if nvidia_controllers:
            print("\nNVIDIA USB 控制器:")
            for controller in nvidia_controllers:
                print(f"- {controller}")
                if "Type-C" in controller:
                    print("  * 支持 USB Type-C")
                    # Check for Thunderbolt support
                    try:
                        tb_check = subprocess.check_output(
                            ["wmic", "path", "Win32_PnPEntity", "where", 
                             "caption like '%Thunderbolt%'", "get", "caption"], 
                            encoding='gbk'
                        )
                        if len(tb_check.strip().split('\n')) > 1:  # Has content beyond header
                            print("  * 支持雷电接口")
                    except:
                        pass
        else:
            print("未找到 NVIDIA USB 控制器")
    except Exception as e:
        print(f"获取 USB 控制器信息失败: {e}")

if __name__ == "__main__":
    get_gpu_info()
    get_usb_controller_info()