When I want to use GPU by trial and error of machine learning, I use AWS because my PC does not have GPU. The way at that time.
environment
Rent a GPU instance
Search AWS Marketplace for "Deep Learning AMI" and select the AMI that appears. (I will use Ubuntu this time)
Select "g2.2xlarge" as the instance type
Instance details settings. Here's the right IAM role. If possible, make it a spot instance and it will be cheaper
Security group settings Change the SSH connection source from the default 0.0.0.0/0 to'My IP'
Create.
Set up SSH transfer
Download the key pair (* .pem) and convert it to a .ppk file with PuTTYgen.
Open [Settings]> [SSH Transfer] from the TeraTerm menu.
From [Add], enter 8888 for [Local Port], localhost for [Remote Host], and 8888 for [Port]. Do not enter anything in [Listen]. (It is important that the remote host is localhost. The transfer destination from the place connected by SSH is the machine itself)
Log in at the terminal
Enter [email protected] in the host name.
Select [RSA / ... Use key] as the authentication method and specify the .ppk file (converted by PuTTY) as the private key.
After logging in, start jupyter.
$ jupyter notebook
View in browser The login URL will appear in the terminal, so enter that URL in your browser to open it.
Recommended Posts