ARTIFICIAL INTELLIGENCE (18) – Deep learning (16) Google Cloud – Compute Engine. Run the AI model on the cloud

An SSH key it’s like a door key that lets you enter the Google Cloud server safely.

It has two parts:

• Public key → You will copy this into Google Cloud.
• Private key → Stays in your computer. Never share it.

 

• Open https://console.cloud.google.com/ and sign in.

• Go to Compute Engine → VM instances.

Add keys to VMs that use OS Login

In the following link you can see the Google instruccions to add the keys:

https://docs.cloud.google.com/compute/docs/connect/add-ssh-keys

How to Add Your SSH Public Key to a Google Cloud VM

When creating a Virtual Machine (VM) in Google Cloud, you often need to add your own SSH public key so you can connect securely from your computer. Here is a simple step‑by‑step explanation of how to do it, assuming you already generated an SSH key pair on your local machine, as explained previously.

1. Open your public SSH key on your computer

Your public key is stored in a file that ends with .pub.
For example, on Windows you can open it using PowerShell:

notepad $env:USERPROFILE\.ssh\aidl2026.pub

A Notepad window will open showing a long line that starts with:

Add the key to the VM configuration in Google Cloud

When creating your VM:

1. Scroll down and click “Management, security, disks, networking, sole tenancy” (advanced settings).
2. Open the Security tab.
3. Look for the section called SSH Keys.
4. Click “Add item”.
5. Paste your entire public key into the box.

Google Cloud will automatically install this key on your VM when it is created.

 

 

Connect to the VM using your private key

Once the VM is running, copy its External IP and run this command from your PowerShell terminal:

ssh -i $env:USERPROFILE\.ssh\aidl2026 your_username@<EXTERNAL_IP>

 

 

Connecting to a Google Cloud VM Using Your SSH Key (After Adding the Key)

Once you have pasted your SSH public key into the SSH Keys section of your Google Cloud VM configuration, you are ready to connect to the machine from your local computer. The steps below describe the full process and what to expect during your first connection.

1. Retrieve the External IP Address of the VM

After the VM is created, open:

Google Cloud Console : Compute Engine : VM Instances

You will see a column named External IP.
This is the address you will use to connect to the server from your computer.

Example:

34.175.3.72

2. Connect to the VM from your local machine

Open PowerShell (on Windows) and run the following command:

ssh -i $env:USERPROFILE\.ssh\aidl2026 <username>@<EXTERNAL_IP>

Replace:

• <username> with the username you embedded in your SSH public key (the text after the key, e.g., yolanda)
• <EXTERNAL_IP> with the actual external IP of your VM

Example:

ssh -i $env:USERPROFILE\.ssh\aidl2026 yolanda@IPNUMBER

3. Accept the host authenticity prompt

During the first connection, SSH will display a message like:

The authenticity of host ‘IPNUMBER’ can’t be established. Are you sure you want to continue connecting (yes/no/[fingerprint])?

This is completely normal.
Type:

yes

and press Enter.

SSH will then store the server’s fingerprint on your machine.

4. Successful connection

If your SSH key is installed correctly and the VM is running, you will be logged into the server and see a prompt like this:

yolanda@instance-mlops-lab:~$

At this point, you are fully connected to your Google Cloud Virtual Machine over SSH.
You can now execute commands inside the remote Linux environment, install tools, transfer files, configure environments, or continue with any cloud computing or MLOps tasks.

This marks the completion of the secure remote access setup.

you are already inside your Google Cloud VM, with the prompt:

yolanda@instance-mlops-lab:~$

Install Miniconda on the Remote VM

You are now inside a Debian Linux VM.
Run the following commands inside the VM terminal (not in PowerShell).

1. Download the Miniconda installer

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

 

2. Make the installer executable

chmod +x Miniconda3-latest-Linux-x86_64.sh

3. Run the installer

sh Miniconda3-latest-Linux-x86_64.sh

4. Reload your shell

After installation finishes, run:

source ~/.bashrc

Create a new conda environment for the lab

The lab recommends creating a clean environment.
You can create it like this:

conda create -n mlops python=3.10 -y

CondaTSNonInteractiveError: Terms of Service have not been accepted…

This is a very recent change in Conda/Anaconda:
you must accept the Terms of Service (TOS) for the official repo.anaconda.com channels before creating an environment.

Accept Conda Terms of Service

Run these two commands inside your VM:

conda tos accept –override-channels –channel https://repo.anaconda.com/pkgs/main

conda tos accept –override-channels –channel https://repo.anaconda.com/pkgs/r

Create your environment again

Now repeat:

conda create -n mlops python=3.10 -y

Activate the Environment

Once the environment is created, activate it:

conda activate mlops

Your prompt should now change to:

(mlops) yolanda@instance-mlops-lab:~$

This means the environment is active and ready.

This means:

• Miniconda is installed
• The environment mlops exists
• The environment is active

Install the correct PyTorch (CPU‑only)

pip install torch torchvision –index-url https://download.pytorch.org/whl/cpu

Then install the rest:

pip install numpy pandas matplotlib scikit-learn

Move Your Dataset and Code to the VM. Moving the code and the dataset to the server.

Let’s do this step now.

From your local Windows machine, upload your project files into the VM

Open PowerShell on your Windows computer (NOT inside SSH).

• You must upload the folder that contains the images you downloaded (cars vs flowers)
• You must upload the Python files you are going to run (training script, dataset loader, etc.)

These files must be copied from your computer : into the VM.

How to Upload Your Dataset Folder to the VM 

Goal:

Copy the folder cars_vs_flowers (located on your Windows Desktop) into your Google Cloud VM.

To do this, you must use the scp command from your local computer, not from inside the VM.

Follow these steps carefully:

Exit the VM (if you are inside SSH)

If your terminal shows something like this:

(mlops) yolanda@instance-mlops-lab:~$

you are inside the VM.

Type:

exit

This means you are back on your Windows machine, which is where scp must run.

Upload the folder from Windows using SCP

Now, from PowerShell on your Windows PC (NOT inside SSH), run:

scp -i $env:USERPROFILE\.ssh\aidl2026 -r «C:\Users\yolanda.muriel\Desktop\cars_vs_flowers» yolanda@IPnumber:~

This is the version that includes:

• -i → identity file (your private key)
• full path to the folder
• your username
• your VM IP

This is EXACTLY like your SSH command — just with scp.

To upload a folder:

scp -r «C:\Users\yolanda.muriel\Desktop\cars_vs_flowers» yolanda@IPNUMBER2:~

Why it works

Your SSH login uses:

ssh -i $env:USERPROFILE\.ssh\aidl2026 yolanda@IPNUMBER

So SCP must also include -i : same key.

Without the key:

• SCP tries to log in with a password
• your VM does not allow passwords
• result: Permission denied (publickey)

Confirm the dataset is inside the VM

Now that the upload succeeded, connect back to your VM:

ssh -i $env:USERPROFILE\.ssh\aidl2026 yolanda@NUMBERIP

Then, inside the VM:

ls

Open or edit the project remotely

• Option A — Continue working directly inside the SSH terminal, or
• Option B — Use VSCode Remote SSH (recommended)

Using VSCode Remote SSH to Work on the VM (Recommended)

The easiest way to edit your code on the remote VM is by using Visual Studio Code with the Remote – SSH extension. This allows you to open files, edit Python scripts, run training code, and manage your project as if it were on your own computer.

1. Install the Remote SSH extension in VSCode

1. Open Visual Studio Code
2. Go to Extensions
3. Search for: Remote – SSH
4. Click Install

 

REMOTE‑SSH EXTENSION IS INSTALLED SUCCESSFULLY. The next step is connecting your VS Code to your Google Cloud VM.

STEP 1 — Add your VM as an SSH host in VS Code

Open VS Code.

CTRL + SHIFT + P

Once open Command Palette, type:

Remote-SSH: Add New SSH Host

Paste the line:

ssh -i C:\Users\yolanda.muriel\.ssh\aidl2026 yolanda@IPNUMBER

 

Then VS Code asks:

Where do you want to save the SSH configuration?

Choose:

C:\Users\yolanda.muriel\.ssh\config

(or whichever config it offers inside your .ssh folder)

STEP 2 — Connect to the VM

Remote-SSH: Connect to Host…

Choose:

yolanda@ipnumber

You can also host a different or customized .ssh:

VS Code will now:

• open a NEW WINDOW
• install the VS Code server on your VM
• connect your editor directly to /home/yolanda/ on the VM

If a message appears saying:

“Do you trust the files in this folder?”

Choose YES.

STEP 3 — Open your project folder

Inside VS Code (connected to the VM):

1. Go to File → Open Folder
2. Select:

/home/yolanda/

You should now see your folder:

cars_vs_flowers

and any .py scripts you upload.

 

His is your home directory inside the virtual machine, and inside that folder you will find:

• cars_vs_flowers/ (your dataset)
• your Python scripts (once you upload them)
• miniconda3/
• everything you need for the lab

You are now working directly inside your cloud VM

This means you can:

• edit Python files on the VM
• run training scripts
• upload more files
• debug remotely
• run code inside your conda environment

Everything you do is now happening in the cloud, not on your PC.

Edit Files, Debug, and Train Remotely

Now that your VM’s folder is open inside VS Code, you can work exactly as if the files were on your computer. This is the moment where you can do:

• Edit your Python training script
  Load the dataset you uploaded
  Run the training through the VS Code terminal
  Debug remotely inside the cloud VM

Here is how to proceed.

1. Open the VS Code Terminal inside the VM

In VS Code, press:

CTRL + SHIFT + `

or use:

Terminal : New Terminal

IMPORTANT:
This terminal is not your PC —
it is the terminal inside the Google Cloud VM, in the folder:

/home/yolanda/

2. Activate your Conda environment

Inside the VS Code terminal:

conda activate mlops

You should see:

(mlops) yolanda@instance-mlops-lab:~$

This means all your installed libraries (torch, numpy, pandas…) are available.

Navigate to your project folder

If your scripts are inside your home directory:

cd /home/yolanda/

Prepare Your Dataset Loader

Now that your dataset folders are in the correct place, you can do:

• Load the dataset using PyTorch’s ImageFolder
• Point it to the correct path inside the VM
• Start training the model

1. Confirm dataset path in your training script

In the VM, your dataset path is:

/home/yolanda/cars_vs_flowers/training_set

So in your Python code, use:

STEP 1 — Create or open your training script

Inside VS Code (connected to the VM):

1. In the Explorer, choose the folder where you want your code (for example /home/yolanda/).
2. Right‑click → New File
3. Name it:

train.py

This is where you paste the dataset‑loading code.

STEP 2 — Paste the dataset loader inside that file

Here is the code again:

Run it in the VM terminal (inside VS Code)

Open terminal:

CTRL + SHIFT + `

Activate your conda environment:

conda activate mlops

Then run your script:

python train.py

That confirms the dataset loads correctly from the VM.
Now let’s add a simple training pipeline (model + loss + optimizer + loop) so you can complete the  end‑to‑end.

Complete training script (train.py)

 

@Yolanda Muriel Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0)