Ollama Guide

Guide to Using Ollama

Ollama allows you to run large language models (LLMs) like LLaMA, Mistral, and more on your local machine.

---

1️⃣ Install Ollama

🔹 On macOS & Linux

Run the following command in your terminal:

curl -fsSL https://ollama.com/install.sh | sh

🔹 On Windows

1. Download Ollama from Ollama’s official website.
2. Run the installer and follow the setup instructions.

---

2️⃣ Running Ollama

✅ Check if Ollama is installed

ollama --version

📜 List available models

ollama list

📥 Download and Run a Model

To run Llama 2:

ollama run llama2

To run Mistral:

ollama run mistral

---

3️⃣ Using Ollama in a Python Script

🛠 Install Ollama’s Python package

pip install ollama

💻 Run a simple Python script

import ollama

response = ollama.chat(model='mistral', messages=[{'role': 'user', 'content': 'What is Ollama?'}])
print(response['message']['content'])

---

4️⃣ Customizing Models

You can create your own fine-tuned models with Modelfiles.

📄 Create a custom model

1. Create a Modelfile:

   FROM mistral
   

2. Run:

   ollama create mymodel -f Modelfile
   

3. Use it with:

   ollama run mymodel
   

---

5️⃣ Running Ollama as an API

Ollama provides an API to integrate with your applications.

🖥 Start the Ollama server

ollama serve

📡 Send API requests using curl

curl http://localhost:11434/api/generate -d '{ "model": "mistral", "prompt": "Hello, how are you?" }'

---

🚀 Final Thoughts

• Ollama is great for running LLMs locally without sending data to the cloud.
• It works offline once the models are downloaded.
• You can fine-tune and customize models for specific tasks.

Would you like help setting up a specific model or using Ollama for a project? 😊

🚀 Creating Your Own Model in Ollama

Ollama allows you to customize and fine-tune models using Modelfile.

---

1️⃣ Install Ollama

If you haven’t installed Ollama yet, do so first:

curl -fsSL https://ollama.com/install.sh | sh

Check if it’s installed:

ollama --version

---

2️⃣ Create a Custom Model Using a Modelfile

A Modelfile is a simple way to customize or fine-tune models in Ollama.

📄 Step 1: Create a Modelfile

Create a new file named Modelfile:

touch Modelfile

Open it in an editor:

nano Modelfile

✍️ Step 2: Define Your Model in the Modelfile

Here’s an example of a custom model using Mistral as a base:

FROM mistral

# Add system instructions
SYSTEM "You are an AI assistant trained for research tasks."

# Add example prompts & responses to guide the model
# Helps improve fine-tuning for specific applications
PROMPT "Translate the following English text to French:"
RESPONSE "Bonjour! Comment ça va?"

---

3️⃣ Build Your Custom Model

After defining your Modelfile, run:

ollama create mymodel -f Modelfile

This will create a new custom model named “mymodel”.

---

4️⃣ Run Your Custom Model

You can now use your fine-tuned model by running:

ollama run mymodel

Or interact via API:

curl http://localhost:11434/api/generate -d '{ "model": "mymodel", "prompt": "Hello!" }'

---

5️⃣ (Optional) Modify and Improve Your Model

You can enhance your Modelfile by:

• Adding more instructions to change behavior.
• Including specific prompt-response pairs for improved accuracy.
• Using different base models like llama2, gemma, or mistral.

Example:

FROM llama2

SYSTEM "You are an AI specialized in medical diagnosis."

PROMPT "Describe symptoms of a common cold."
RESPONSE "The common cold often includes sneezing, runny nose, and mild fever."

Rebuild after changes:

ollama create newmodel -f Modelfile

---

🎯 Conclusion

• Custom models let you fine-tune behavior for specific tasks.
• Modelfiles help define instructions and sample responses.
• You can build and deploy models locally without cloud dependency.

🚀 Now you have a custom AI model in Ollama! 🎉 Let me know if you need further help!