Access AWS from local machine

First and foremost, add AWS access keys to env variables to local machine, ideally in ~/.zsh (or whatver your $SHELL is).

export AWS_ACCESS_KEY_ID="..."

these can be found in the AWS console under IAM -> Users -> Security credentials.

Start/stop an instance

aws ec2 start-instances --instance-ids i-0a92dcb7d252071c5
aws ec2 stop-instances --instance-ids i-0a92dcb7d252071c5

or via IP address:

aws ec2 start-instances --instance-ids $(aws ec2 describe-instances --filters "Name=ip-address,Values=" --query 'Reservations[*].Instances[*].InstanceId' --output text)
aws ec2 stop-instances --instance-ids $(aws ec2 describe-instances --filters "Name=ip-address,Values=" --query 'Reservations[*].Instances[*].InstanceId' --output text)

ssh and connect

Optionally, create and assign elastic IP to the instance (cost is $3.5 per month) for easier access.

If you have a pem key you can just log in with:

ssh -i my-aws-gpu-machine.pem ubuntu@

or if you make a ~/.ssh/config file like this:

Host my-aws-gpu-instance 
  User ubuntu
  IdentityFile ~/.ssh/my-aws-gpu-instance.pem   # if you lose this you can use the trick below

then just ssh my-aws-gpu-instance.

If you happen to lose pem file, just: - create a key pair run this on LOCAL machine (skip this if already exists and just copy it: cat ~/.ssh/id_rsa.pub) bash ssh-keygen -t rsa -b 4096 -C "<pick a name of a machine you are running this command on>" - log into remote machine using ec2 instance connect - copy LOCAL public key to authorized_keys: echo public_key_string >> ~/.ssh/authorized_keys or use nano editor. - change permissions on REMOTE machine: bash chmod 700 ~/.ssh chmod 600 ~/.ssh/authorized_keys

Change shell prompt appearance

Open ~/.bashrc and add the following lines:

# Function to get public IP address
get_public_ip() {
    curl -s ifconfig.me

# Set the PROMPT variable to include the public IP address
PROMPT='%n@$(get_public_ip) %~ %# '

Access via VSCode

On the left just add remote instance, or Cmd+Shift+P and type Remote-SSH: Connect to Host... and type ssh my-aws-gpu-instance.

Add github access

Take a public key from any machine and copy/paste into GitHub - Settings - SSH and GPG keys - New SSH key.

Anything GPU

To see recommended drivers:

sudo apt install ubuntu-drivers-common
ubuntu-drivers devices

for T4 GPU, the recommended driver is 440. To install the latest driver, run the following commands:

sudo apt update
sudo apt-get upgrade   # conservative upgrade
sudo apt-get dist-upgrade   # aggressive upgrade, be careful
sudo apt install nvidia-driver-545
sudo reboot  # don't forget to reboot, might even need to purge old drivers via `sudo apt-get purge nvidia-driver-XXX`

You can now see stats:

lspci | grep -i nvidia

you should see something like this:

NVIDIA-SMI 545.29.06              Driver Version: 545.29.06    CUDA Version: 12.3    

Install any pyenv and python

You do this only once:

# Install pyenv prerequisites
sudo apt update
sudo apt install -y make build-essential libssl-dev zlib1g-dev libbz2-dev \
libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev \
xz-utils tk-dev libffi-dev liblzma-dev python3-openssl git

# Install pyenv
curl https://pyenv.run | bash

# Add pyenv to the shell
echo 'export PATH="$HOME/.pyenv/bin:$PATH"' >> ~/.bashrc
echo 'eval "$(pyenv init --path)"' >> ~/.bashrc
echo 'eval "$(pyenv virtualenv-init -)"' >> ~/.bashrc
source ~/.bashrc

# Install Python 3.10.14
pyenv install 3.10.14

Set global Python version:

pyenv global 3.10.14

or local:

pyenv local 3.10.14

Also have option to add to the PYTHONPATH:


github username can also come handy:

export GITHUB_USERNAME="nesaboz"

Install AWS CLI

sudo apt install unzip
curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
sudo ./aws/install
source ~/.bashrc
rm awscliv2.zip
aws --version

aws-cli/2.15.41 Python/3.11.8 Linux/6.5.0-1018-aws exe/x86_64.ubuntu.22 prompt/off

Install virtual env in a project

pyenv local 3.10.14  # set up the local 
python -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

to install specific version of a package:

pip install example-package==2.0.1
pip install pip-tools
pip freeze > requirements.txt  # better is to use pip-compile
pip-compile requirements.in
pip-compile --output-file dev.txt requirements.in  # if you want different name

requirements file uses == for versions

Encode time zone

echo "export TZ='PST8PDT'" >> ~/.bashrc

If you must get Conda

Get mamba which is a faster version of conda from here:

wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-Linux-x86_64.sh

or just get miniconda:

wget https://repo.anaconda.com/miniconda/Miniconda3-py310_23.3.1-0-Linux-x86_64.sh

Install it via:

source <file you want to run>.sh

Update mamba/conda:

conda update -n base -c conda-forge conda

Init conda:

conda init zsh  # Must restart terminal if running this one.

Note for Windows

Install WSL for Windows:

wsl --install Ubuntu

will need to add both conda and mamba to PATH by editing ~/.bashrc file and adding these two lines to the end:

source "${CONDA PATH}/etc/profile.d/conda.sh"
source "${CONDA PATH}/etc/profile.d/mamba.sh"

Also to set up default terminal in VSCode follow quick instructions here.

Windows C hard drive will be by default mounted in /mnt/c/ folder.

Use PowerToys/Keyboard Manager to remap the keyboard.

Create environment

Create environment with env.yml or from scratch:

conda create -f env.yml -n custom name # notice there is no keyword env
yes | conda create -n repo_name python=python_ver
conda activate repo_name

Environment management

conda env list  # list all envs
conda env remove -n ENV_NAME
conda list torch  # lists only package version

Install packages

yes | conda install -c conda-forge jupyter_contrib_nbextensions graphviz python-graphviz

Export environment

Alternatively, export environment to a file:

conda env export > env.yml

or if it is cross-platform export simpler no-build number environment:

conda env export --no-builds | grep -v "prefix" > env.yml

Jupyter kernels

Add a kernel:

ipython kernel install --name $repo_name --user

Using bash script

  • If making shell file make sure it has correct line endings for the system (for example in PyCharm go to File -> File Properties -> Line Separators)

Make sure the file is executable:

chmod +x utils/setup_new_project.sh check with ls -l
./utils/setup_new_project.sh <repo_name> <python_version>
# for example ./setup_new_project.sh test_repo 3.10

To stop execution after a failure use || exit, for example if the folder doesn’t exist following command will stop the script:

cd $repo_name || exit

VS Code Extensions

Python, Jupyter, autoDocstring, Code Spell Checker, Dev Containers.