Skip to content

Using the Jupyter image

For convenience, Leafcloud offers a pre-built image with Nvidia CUDA and CuDNN, TensorFlow and PyTorch and Jupyter Notebook pre-installed.

Launching the instance

To start an instance using this image, go to the instance view of the dashboard.

  1. Press Launch instance
  2. in the Source tab select "Image" and the boot source
  3. select the ubuntu-pytorch-tensorflow-jupyter image
  4. set the volume size to 40GB
  5. as the flavor select any gpu flavor, starting with eg1.
  6. Select a network, to give the instance a public IP address use the external network
  7. Select a key pair, this should will be used to SSH into the machine
  8. Select a security group that exposes the 8888 port
  9. Press "Launch Instance"

Open Jupyter Notebook

In the instances view, copy the IP address of your new instance.

Obtaining the Jupyter Notebook token

From a terminal SSH into the instance: 

ssh ubuntu@<ip-address>
sudo systemctl status jupyter

Near the bottom of the output of the above command will be a line simila to: 

http://45.135.56.154:8888/?token=<your-token>
Copy this url and open it in a web browser.

This is your Jupyter Notebook server!

Validating the installation

To validate that the machine is using a GPU you can create a new notebook and run the following

For TensorFlow

import tensorflow as tf 

if tf.test.gpu_device_name(): 
    print('Default GPU Device:{}'.format(tf.test.gpu_device_name()))
else:
   print("Please install GPU version of TF")

This should output something like: 

Default GPU Device:/device:GPU:0

For PyTorch

import torch
torch.cuda.is_available()
torch.cuda.device_count()
torch.cuda.current_device()
torch.cuda.device(0)
torch.cuda.get_device_name(0)

This should output something like: 

'NVIDIA A30'