Skip to main content

Custom Environment

1. Create a New Instance​

If the platform image does not contain the required versions of Python, Cuda, or frameworks, you can choose Miniconda and then install the environment according to your needs.

new-env-001

2. Install Python​

Log in to the instance terminal and create a virtual environment with the required Python version.

conda create -n gpugeek python==3.8.10
conda activate gpugeek
python3 --version
Python 3.8.10

3. Install Cuda​

Install the required Cuda version.

Go to NVIDIA to download the required CUDA Toolkit installer.

new-env-002

Copy the command from the above image, then download and run it in the terminal. Before downloading, you can use Academic Resource Acceleration.

wget https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda_12.4.0_550.54.14_linux.run
sh cuda_12.4.0_550.54.14_linux.run --silent --toolkit && rm cuda_12.4.0_550.54.14_linux.run

Verify the newly installed Cuda version:

nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Tue_Feb_27_16:19:38_PST_2024
Cuda compilation tools, release 12.4, V12.4.99
Build cuda_12.4.r12.4/compiler.33961263_0

4. Install Frameworks​

Install the required frameworks and versions as needed.

pip config set global.index-url https://mirrors.aliyun.com/pypi/simple
pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu124

5. Install Other Environments​

The versions of Python, Cuda, and frameworks mentioned above should be installed according to your needs. The above steps are only installation examples.

6. Verify Installed Versions​

(gpugeek) root@gz-ins-636197260402693:~# cat check_version.py 
import torch
import sys
x = torch.rand(5, 3)
print("Result:", x)
print("CUDA is available:", torch.cuda.is_available())
print(torch.zeros(1).cuda())
print("GPU available numbers:", torch._C._cuda_getDeviceCount())
print("PyTorch version:", torch.__version__)
print("CUDA version:", torch.version.cuda)
print("Python version:", sys.version)
cudnn_version = torch.backends.cudnn.version()
print(f"cuDNN version: {cudnn_version}")
print("NCCL version:", torch.cuda.nccl.version())


(gpugeek) root@gz-ins-636197260402693:~# python check_version.py 
Result: tensor([[0.7838, 0.1052, 0.2517],
        [0.0549, 0.0639, 0.9170],
        [0.8569, 0.8401, 0.7786],
        [0.2685, 0.7826, 0.3778],
        [0.5491, 0.1513, 0.2379]])
CUDA is available: True
tensor([0.], device='cuda:0')
GPU available numbers: 1
PyTorch version: 2.4.0+cu124
CUDA version: 12.4
Python version: 3.8.10 (default, Jun  4 2021, 15:09:15) 
[GCC 7.5.0]
cuDNN version: 90100
NCCL version: (2, 20, 5)