Pytorch 安装环境配置 (old)

建立环境

1. 建立一个新环境 name = pytorch

   conda create -n pytorch python

2. 进入 pytorch 工作环境 conda activate pytorch

3. 查看安装包 pip list

4. 查询显卡型号:

   任务管理器中，查看 Geforce GTX 950M https://developer.nvidia.com/cuda-gpus

   查看驱动版本： C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi.exe

   Driver Version: 388.57 ==>需要升级驱动 https://developer.nvidia.com/drive/downloads

查看驱动版本：check graphic card and driver both suppot the CUDA varsion

cd C:\Program Files\NVIDIA Corporation\NVSMIconda //windows

nvidia-smi.exe

   NVIDIA-SMI version: 512.15
   Card type: NVIDIA GeForce 
   CUDA Version: 11.6

Check Compute Capabilities:

https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

### CUDA Toolkit version need match Card Driver Version 
	
	CUDA 12.0.x     >=527.41 (Windows Driver)
	CUDA 11.6.x     >=452.39 (Windows Driver)

| NVIDIA-SMI 512.15 Driver Version: 512.15 === CUDA Version: 11.6 |

安装CUDA Toolkit and NVIDIA Driver

NVIDIA Driver Downloads: https://www.nvidia.com/download/index.aspx

CUDA Toolkit Downloads: https://developer.nvidia.com/cuda-downloads

安装 pytorch

1. 选择系统，软件包，下载版本，产生下载安装命令 https://pytorch.org/get-started/locally/

Anaconda Prompt (anaconda3)

conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia

Why pip install does not working?

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117

查询是否有torch

// Run Anaconda Prompt 
 (base) conda activate pytorch      ## activate pytorch env
 (pytorch) python
 >>import torch                    ## 查看是否可以使用GPU 
 >>torch.cuda.is_available() 
    True
 >>torch.version.cuda
    11.6

The solution of GPU is not available (if torch.version.cuda return none)

• Check CUDA Toolkit version match driver version

• conda uninstall pytorch

• conda uninstall cpuonly

• select to install pytroch v1.14 or v2.0

  • conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia (v1.14)

  • https://pytorch.org/get-started/pytorch-2.0/

Run Jupyter lab with GPU

Method 1: Run Anaconda Navigator, Launch JupyterLab

Method 2: Run Anaconda Prompt (Anaconda3) -- entry command, then run jupyterlab.bat jupyterlab.bat: D:\ProgramData\Anaconda3\python.exe d:\ProgramData\Anaconda3\cwp.py d:\ProgramData\Anaconda3 d:\ProgramData\Anaconda3\python.exe d:\ProgramData\Anaconda3\Scripts\jupyter-lab-script.py "%USERPROFILE%/"

Install pytorch genmetric https://pytorch-geometric.readthedocs.io/en/latest/install/installation.html

pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.13.0+cu117.html

torch.cuda.device_count()      // 
torch.cuda.current_device() 
torch.cuda.device(0) 
torch.cuda.get_device_name(0)
torch.zeros(1).cuda()      // 

# setting device on GPU if available, else CPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)
print()

#Additional Info when using cuda
if device.type == 'cuda':
    print(torch.cuda.get_device_name(0))
    print('Memory Usage:')
    print('Allocated:', round(torch.cuda.memory_allocated(0)/1024**3,1), 'GB')
    print('Cached:   ', round(torch.cuda.memory_reserved(0)/1024**3,1), 'GB')

## check torch version //v1.11

print(torch.version) // in jupyter python -c "import torch; print(torch.version)"

##To move tensors to the respective device: torch.rand(10).to(device)

##To create a tensor directly on the device: torch.rand(10, device=device)

安装其他包

// install torch geometric

   refer to:  https://pytorch-geometric.readthedocs.io/en/latest/notes/installation.html
   conda install pyg -c pyg    //conda
   pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.11.0+cu115.html

// install jurpyter in pytorch enveriment
   conda list                //check installed parkage
   conda install jupyter

Does PyTorch see any GPUs?		torch.cuda.is_available()
Are tensors stored on GPU by default?	torch.rand(10).device
Set default tensor type to CUDA:	torch.set_default_tensor_type(torch.cuda.FloatTensor)
Is this tensor a GPU tensor?		my_tensor.is_cuda
Is this model stored on the GPU?	all(p.is_cuda for p in my_model.parameters())