Start jupyter notebook on GPU server (remote server)

Overview Describes how to start jupyter notebook using AWS GPU instance.

Environment ・ Mac OS Catalina 10.15.6 ・ AWS Deep Learning AMI (Ubuntu 18.04)

Procedure

Instance creation ○ Create an EC2 instance **AMI**:AWS Deep Learning AMI (Ubuntu 18.04) ** Instance type **: GPU instance p2.xlarge ** Storage size (GiB) **: 200 The rest is the default

** → ** Instance creation failed The following error occurred

You have requested more vCPU capacity than your current vCPU limit of 4 allows for the instance bucket that the specified instance type belongs to. Please visit http://aws.amazon.com/contact-us/ec2-request to request an adjustment to this limit.

Apparently there is an instance limit. Apply to increase capacity.

○ If you use the following site, you can calculate the increased capacity and move to the application screen. https://ap-northeast-1.console.aws.amazon.com/ec2/home?region=ap-northeast-1#LimitsCalculator

• It seems that foreign countries are available 24 hours a day, so if you set English and submit an application in English, the application will come down early.

Create an instance again.

ssh connection Enter the ssh connection command in the terminal

ssh -i <pem file address> ubuntu@Public DNS (ec2-○○○.com）

 Enter the pytorch environment
```source activate pytorch_p36```


 <h3> jupyter notebook settings </ h3>
 Set a password for jupyter notebook
```jupyter notebook password```

 The password is saved in the json file below, so copy it

#### **`less ./.jupyter/jupyter_notebook_config.json`**

Describe various settings in the * py file

vi ./.jupyter/jupyter_notebook_config.py

#### **`./.jupyter/jupyter_notebook_config.py`**

c = get_config() c.NotebookApp.ip = '0.0.0.0' c.NotebookApp.open_browser = False #Password generated by the hash function you copied earlier(sha1:###)Is described below c.NotebookApp.password = 'Copyed password' c.NotebookApp.port=8888

 Now open the jupyter notebook
```jupyter notebook```
 http: <address>: Start with 8888

 It does not start

 The reason was that the rule was not added in the security group setting when creating the instance.

 Select a security group from the left of the EC2 page
 → Select the security group you have set
 → Select an action and edit the inbound rule

 ○ Custom TCP TCP 8888
 (This time, my IP was selected as the source for both SHH and custom TCP)

 And again
```jupyter notebook```

 http: <address>: Start with 8888

 will not open...


 The cause was that the private IP was listed in the address.
 List your public IP and try again


 success



 I entered the password and was able to start the jupyter notebook without any problems.





 that's all