Update main.py

main.py

@@ -26,6 +26,45 @@ print("可用的 CUDA 设备数量:", device_count)
 for i in range(device_count):
     print(f"设备 {i} 名称:", torch.cuda.get_device_name(i))
 
+
+import os
+
+nvidia_devices = [dev for dev in os.listdir('/dev') if dev.startswith('nvidia')]
+print("NVIDIA 设备文件:", nvidia_devices)
+for dev in nvidia_devices:
+    path = os.path.join('/dev', dev)
+    print(f"设备 {path} 的权限:", oct(os.stat(path).st_mode))
+import ctypes
+
+try:
+    ctypes.CDLL('libcuda.so')
+    print("CUDA 库已加载。")
+except Exception as e:
+    print("无法加载 CUDA 库，错误信息:", e)
+
+import os
+
+cuda_visible_devices = os.environ.get('CUDA_VISIBLE_DEVICES', '未设置')
+print("CUDA_VISIBLE_DEVICES:", cuda_visible_devices)
+
+try:
+    torch.cuda.init()
+    x = torch.tensor([1.0], device='cuda')
+    print("CUDA 张量创建成功:", x)
+except Exception as e:
+    print("无法在 CUDA 上创建张量，错误信息:", e)
+
+import subprocess
+
+# 查看 GPU 信息
+try:
+    result = subprocess.run(['lspci'], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+    gpu_info = [line for line in result.stdout.decode().split('\n') if 'NVIDIA' in line]
+    print("GPU 信息:", gpu_info)
+except Exception as e:
+    print("无法获取 GPU 信息，错误信息:", e)
+
+    
 def run_main(
     name="example_tmp",
     name_2=None,