deepseek

deepseek

Tags
Model
Price, $ per user

Reading list
👨‍🔧 Use cases

Deepseek

DeepSeek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese. We provide various sizes of the code model, ranging from 1B to 33B versions. Each model is pre-trained on project-level code corpus by employing a window size of 16K and an extra fill-in-the-blank task, to support project-level code completion and infilling. For coding capabilities, DeepSeek Coder achieves state-of-the-art performance among open-source code models on multiple programming languages and various benchmarks.

  • Massive Training Data: Trained from scratch on 2T tokens, including 87% code and 13% linguistic data in both English and Chinese languages.
  • Highly Flexible & Scalable: Offered in model sizes of 1B, 5.7B, 6.7B and 33B, enabling users to choose the setup most suitable for their requirements.
  • Superior Model Performance: State-of-the-art performance among publicly available code models on HumanEval, MultiPL-E, MBPP, DS-1000, and APPS benchmarks.
  • Advanced Code Completion Capabilities: A window size of 16K and a fill-in-the-blank task, supporting project-level code completion and infilling tasks.

Comparison with other models

image
image

Feedback

📝
Greetings, it appears that no one has yet discussed https://chat.deepseek.com/coder. I transitioned to it after using Phind. It comprises two models - a general one and an instruct model. These open-source models can be implemented locally in Ollama. Overall, it operates swiftly. I use the instruct model primarily for the following tasks:
  • Generating a service skeleton based on API contracts/configurations
  • Efficiently extracting necessary information from documentation and generating code based on it
  • Writing scripts to convert from one format to another, such as generating Terraform manifests based on JSON
  • Various bash scripts and commands
  • Writing tests effectively

I use a minimal prompt, only writing "act as <>", the task itself, and "think step by step". In the end, I insert code examples, documentation, and configurations.

In my experience, it reduces routine work time by 2 to 10 times. Recently, I accomplished a task in a couple of days that was estimated for a sprint - transitioning system setting control to Terraform.