WaveCoder-Ultra-6.7B Open-Source Code Large Model - Free Deployment to Solve Problems in the Code Domain

Wavecoder Ultra 6.7b

Developed by microsoft

WaveCoder is a series of large language models for the code domain, designed to solve code-related problems through instruction-following learning. WaveCoder-Ultra-6.7B is the high-performance model in this series.

Large Language Model

Transformers

Supports Multiple LanguagesOpen Source License:MIT #Code Instruction Tuning #Multi-task Code Model #LLM-generated Data

Downloads 92

Release Time : 4/11/2024

Model Overview

WaveCoder aims to solve general code tasks such as code generation, code summarization, code translation, and code repair through instruction tuning. Its training data is generated from a subset of the code-search-net dataset using an LLM generator-discriminator framework.

Model Features

Multi-task Code Processing

Supports four general code tasks: code generation, summarization, translation, and repair.

High-quality Training Data

Refined training data generated from a subset of the code-search-net dataset using an LLM generator-discriminator framework.

High Performance

Excellent performance on benchmarks like HumanEval and MBPP, approaching GPT-4 levels.

Model Capabilities

Code Generation

Code Summarization

Code Translation

Code Repair

Instruction Following

Use Cases

Software Development

Code Autocompletion

Automatically generates code snippets based on context.

Improves development efficiency.

Code Error Fixing

Identifies and fixes errors in code.

Improves code quality.

Code Education

Code Explanation

Provides explanations for complex code.

Helps learners understand code logic.

🚀 🌊 WaveCoder: Widespread And Versatile Enhanced Code LLM

WaveCoder is a series of large language models (LLMs) for the coding domain. It aims to solve relevant problems in the code field through instruction - following learning. The training dataset covers four general code - related tasks: code generation, code summary, code translation, and code repair.

🚀 Quick Start

Please refer to WaveCoder's GitHub repo for inference, evaluation, and training code.

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("microsoft/wavecoder-ultra-6.7b")
model = AutoModelForCausalLM.from_pretrained("microsoft/wavecoder-ultra-6.7b")

✨ Features

WaveCoder 🌊 is a series of large language models (LLMs) for the coding domain, designed to solve relevant problems in the field of code through instruction - following learning. Its training dataset was generated from a subset of code - search - net data using a generator - discriminator framework based on LLMs that the team proposed, covering four general code - related tasks: code generation, code summary, code translation, and code repair.

Model	HumanEval	MBPP(500)	HumanEval Fix(Avg.)	HumanEval Explain(Avg.)
GPT - 4	85.4	-	47.8	52.1
[🌊 WaveCoder - DS - 6.7B](https://huggingface.co/microsoft/wavecoder - ds - 6.7b)	65.8	63.0	49.5	40.8
[🌊 WaveCoder - Pro - 6.7B](https://huggingface.co/microsoft/wavecoder - pro - 6.7b)	74.4	63.4	52.1	43.0
[🌊 WaveCoder - Ultra - 6.7B](https://huggingface.co/microsoft/wavecoder - ultra - 6.7b)	79.9	64.6	52.3	45.7

🔥 News

[2024/04/10] 🔥🔥🔥 WaveCoder repo, models released at [🤗 HuggingFace](https://huggingface.co/microsoft/wavecoder - ultra - 6.7b)!
[2023/12/26] WaveCoder paper released.

📄 License

This code repository is licensed under the MIT License. The use of DeepSeek Coder models is subject to its [License](https://github.com/deepseek - ai/DeepSeek - Coder/blob/main/LICENSE - MODEL).

☕️ Citation

If you find this repository helpful, please consider citing our paper:

@article{yu2023wavecoder,
  title={Wavecoder: Widespread and versatile enhanced instruction tuning with refined data generation},
  author={Yu, Zhaojian and Zhang, Xin and Shang, Ning and Huang, Yangyu and Xu, Can and Zhao, Yishujie and Hu, Wenxiang and Yin, Qiufeng},
  journal={arXiv preprint arXiv:2312.14187},
  year={2023}
}

⚠️ Important Note

WaveCoder models are trained on the synthetic data generated by OpenAI models. Please pay attention to OpenAI's [terms of use](https://openai.com/policies/terms - of - use) when using the models and the datasets.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご