DeepSeek Coder for Ollama

DeepSeek Coder offers unparalleled coding capabilities and flexibility, addressing a wide range of programming requirements.

Introduction

DeepSeek Coder is a groundbreaking series of models, each trained from scratch on a massive 2T token dataset, comprised of 87% code and 13% natural language in both English and Chinese. These models are tailored for project-level code completion and infilling, showcasing state-of-the-art performance in various programming languages.

Key Features

• Massive Training Data: Built from 2T tokens with a blend of 87% code and 13% linguistic data in English and Chinese.
  
• Model Variants: Available in sizes of 1.3B, 5.7B, 6.7B, and 33B.
  
• Top-tier Performance: Leads in benchmarks like HumanEval, MultiPL-E, MBPP, DS-1000, and APPS.
  
• Advanced Code Completion: Employs a 16K window size and a unique fill-in-the-blank task.
  

Models on Ollama

Our repository offers two model variants on the Ollama platform:

• 6.7B Model
  
• 33B Model
  

Model Sources

The DeepSeek models leverage GGUF weight files, which can be found at:

• 6.7B Model Source
  
• 33B Model Source
  

Usage

CLI

Run DeepSeek Coder via CLI:

ollama run g1ibby/deepseek:6.7b "Write me a function that outputs the fibonacci sequence"

License

DeepSeek Coder is licensed under the terms found here.

For more information, visit repository.