2️⃣在Python虚拟环境下使用 VS Code or PyCharm

Implementing PyTorch or PyCharm with VS code in virtual python environment

Check NVIDIA Driver and CUDA Version

1. cmd

2. nvidia-smi

VS Code

1. Install VS Code

2. Open VS Code

3. Install extention (Python, Jupyter)

4. Ctrl + Shift + P

5. Python: Select Interpreter (Select virtual python environment)

6. Select pytorch (Python 3.11.9)

PyCharm

• Install PyCharm

• Create a new project

• Select Custom environment

• Select existing

• Select Conda Type

• Select Environment

Check version and GPU (batch file)

## check_v.py

import torch
import sys

print("Python Version is",sys.version)
print("CUDA Version is", torch.version.cuda)
print("Pytorch Version is", torch.__version__)
print("Whether CUDA is supported by our system:", torch.cuda.is_available())

device_id =  torch.cuda.current_device()
print("Current Device name:", torch.cuda.get_device_name(device_id))

cuda_id = "GPU available for cuda:"+str(device_id)
device = cuda_id if torch.cuda.is_available() else 'Only CPU can be used'
print(device)

## entryTorch.bat

call conda activate pytorch
python check_v.py

Install cuDNN

1. https://developer.nvidia.com/cudnn
2. Download cuDNN Library

3. Select Windows -> x86_64 -> Tarball -> 12 (CUDA Version)

4. Download and Unzip

5. go to C:\Program Files\NVIDIA Corporation

6. Create new folder: CUDNN

7. Create a new folder, name it is v9.0 which in the folder CUDNN

8. copy unzip files into the v9.x folder

9. add path C:\Program Files\NVIDIA Corporation\CUDNN\v9.x\bin into Environment Variables