DeepSeek-R1_distilled

A lightweight Python application hosting deepseek-r1 Large Language Model ran locally.

Requirements

• Python 3.7+
• Libraries: customtkinter, ollama, pillow
• Ollama (for handling the model)
• Deepseek-r1 Reasoning Model

Installation

1. Clone the repository:

   git clone https://github.com/PETROUNKNOWN/DeepSeek-R1_distilled.git
   

2. Install the required libraries:

   pip install customtkinter ollama pillow
   

3. Ollama:

   • Download Ollama from Ollama.com
   • macOS
     • Open the .dmg file you downloaded.
     • Drag and drop the Ollama app icon into the Applications folder.
     • Open Finder, go to Applications, and double-click the Ollama app to launch it.
   • Windows
     • Double-click the .exe file you downloaded.
     • Follow the on-screen instructions in the installation wizard.
     • Choose the installation location (default is fine for most users).
     • Once the installation is complete, find Ollama in the Start Menu or Desktop and open it.
   • Linux
     • Open a terminal.
     • Navigate to the directory where the downloaded file is located.
     • Run the appropriate command based on the file type:

       sudo dpkg -i ollama.deb
       sudo apt-get install -f

     • For .rpm (Fedora/RedHat-based systems):

       sudo rpm -ivh ollama.rpm

     • Once installed, launch Ollama from the Applications menu or by running ollama in the terminal.
   • In a terminal:

     ollama serve

   • An Ollama http instance should now be running on port 11434:
     • Visiting localhost:11434 will take you to a page that says Ollama is running

4. Model deepseek-r1:

   • In a teminal:

     ollama run deepseek-r1:8b

Note

• The model requires a lot of compute hence a dedicated GPU is highly recommended
• The Model download requires 4.9GB of disk space although one could go with a 20.0GB one.
  Instead of ollama run deepseek-r1:8b use ollama run deepseek-r1:32b
• You are recommended to search online other things Ollama can do including other models you can run instead of DeepSeek's deepseek-r1 Model.