Chat with DeepSeek-R1 Models using Ollama on Colab

🔮 Overview

Inside the notebook, you will find:

1. Environment Setup: Get Ollama and required dependencies
2. GPU Check: Ensures you are running on a GPU-enabled runtime (T4 is typical in Colab)
3. Model Download: Choose from five different DeepSeek R1 models based on your GPU capacity
4. Model Listing: View currently installed models
5. Chat Interface: Interact with your selected DeepSeek model directly in Colab

⚡ Features

• Simple Setup: Run a few cells in sequence to have all dependencies ready
• Multiple Model Variants: Choose from 1.5B, 7B, 8B, 14B, or 32B parameters
• Responsive Chat: A convenient prompt cell to chat with DeepSeek-R1

🚀 Quick Start

1. 
2. Execute cells in order

📊 Available Models

Model	Size	Notes
⚡ 1.5b	~1.1 GB	Fastest and lightest
⚡ 7b	~4.8 GB	Balanced performance
⚡ 8b	~4.9 GB	Fast with recommended performance
🐢 14b	~9.0 GB	Slower but more capable
🐢 32b	~20 GB	Most capable, requires large GPU RAM

After selecting a model size, it will download (if not already present) and prepare for inference.

🚨 Limitations

Runtime Disconnections and Storage

• Free Google Colab GPU runtime typically lasts up to 1 hour 20 minutes
• Available disk storage is ~112.6 GB (usage varies based on installed models)
• Runtime automatically disconnects after the time limit or period of inactivity

GPU Availability

• T4 GPU is typically provided for free-tier users
• GPU access may be limited during high-demand periods
• Change runtime: Runtime → Change Runtime Type → GPU → Save

Model Performance

• Larger models (32b) may not run efficiently on free Colab instances
• Recommended models for free accounts: 1.5b and 8b

---

Happy DeepSeeking! 🚀