report-review

Unnamed repository; edit this file 'description' to name the repository.
Log | Files | Refs | Submodules | README | LICENSE

gpu_check.py (997B)


      1 import torch
      2 
      3 # 1. CUDAが利用可能か(CUDA版PyTorchがインストールされているか)
      4 is_available = torch.cuda.is_available()
      5 print(f"torch.cuda.is_available(): {is_available}")
      6 
      7 if is_available:
      8     # 2. PCに認識されているGPUの数
      9     device_count = torch.cuda.device_count()
     10     print(f"torch.cuda.device_count(): {device_count}")
     11 
     12     # 3. 現在使用中のGPUの名前(通常0番目のGPU)
     13     device_name = torch.cuda.get_device_name(0)
     14     print(f"GPU Name: {device_name}")
     15 
     16     # 4. GPU上でテンソル演算を実行できるか
     17     try:
     18         # GPU (CUDA) デバイスを指定
     19         device = torch.device("cuda")
     20         # テンソルをGPUメモリに作成し、演算を実行
     21         tensor_on_gpu = torch.rand(2, 3, device=device)
     22         print("\nSuccessfully created and calculated a tensor on GPU:")
     23         print(tensor_on_gpu)
     24     except Exception as e:
     25         print(f"\nError accessing GPU for calculation: {e}")