zjy2001 commited on
Commit
0ff0082
1 Parent(s): 9f4a26e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +66 -1
README.md CHANGED
@@ -1,3 +1,68 @@
1
  ---
2
- license: mit
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: other
3
+ library_name: transformers
4
+ datasets:
5
+ - humaneval
6
+ license_name: deepseek
7
+ pipeline_tag: text-generation
8
+ tags:
9
+ - code
10
+ metrics:
11
+ - code_eval
12
  ---
13
+ <h1 align="center">
14
+ 🌊 WaveCoder: Widespread And Versatile Enhanced Code LLM
15
+ </h1>
16
+
17
+
18
+ <p align="center">
19
+ <a href="https://arxiv.org/abs/2312.14187"><b>[📜 Paper]</b></a> •
20
+ <!-- <a href=""><b>[🤗 HF Models]</b></a> • -->
21
+ <a href="https://github.com/microsoft/WaveCoder"><b>[🐱 GitHub]</b></a>
22
+ <br>
23
+ <a href="https://twitter.com/TeamCodeLLM_AI"><b>[🐦 Twitter]</b></a> •
24
+ <a href="https://www.reddit.com/r/LocalLLaMA/comments/19a1scy/wavecoderultra67b_claims_to_be_the_2nd_best_model/"><b>[💬 Reddit]</b></a> •
25
+ <a href="https://www.analyticsvidhya.com/blog/2024/01/microsofts-wavecoder-and-codeocean-revolutionize-instruction-tuning/">[🍀 Unofficial Blog]</a>
26
+ <!-- <a href="#-quick-start">Quick Start</a> • -->
27
+ <!-- <a href="#%EF%B8%8F-citation">Citation</a> -->
28
+ </p>
29
+
30
+ <p align="center">
31
+ Repo for "<a href="https://arxiv.org/abs/2312.14187" target="_blank">WaveCoder: Widespread And Versatile Enhanced Instruction Tuning with Refined Data Generation</a>"
32
+ </p>
33
+
34
+
35
+ ## 🔥 News
36
+
37
+ - [2024/04/10] 🔥🔥🔥 WaveCoder repo, models released at [🤗 HuggingFace](https://huggingface.co/microsoft/wavecoder-ultra-6.7b)!
38
+ - [2023/12/26] WaveCoder paper released.
39
+ ## 💡 Introduction
40
+
41
+ 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 we proposed, covering four general code-related tasks: code generation, code summary, code translation, and code repair.
42
+
43
+ | Model | HumanEval | MBPP(500) | HumanEval<br>Fix(Avg.) | HumanEval<br>Explain(Avg.)|
44
+ |---|---|---|---|---|
45
+ | GPT-4 | 85.4 | - | 47.8 | 52.1 |
46
+ | [🌊 WaveCoder-DS-6.7B](https://huggingface.co/microsoft/wavecoder-ds-6.7b) | 65.8 | 63.0 | 49.5 | 40.8|
47
+ | [🌊 WaveCoder-Pro-6.7B](https://huggingface.co/microsoft/wavecoder-pro-6.7b) | 74. 4 | 63.4 | 52.1 | 43.0 |
48
+ | [🌊 WaveCoder-Ultra-6.7B](https://huggingface.co/microsoft/wavecoder-ultra-6.7b) | 79.9 | 64.6 | 52.3 | 45.7 |
49
+
50
+ ## 🪁 Evaluation
51
+
52
+ Please refer to WaveCoder's [GitHub repo](https://github.com/microsoft/WaveCoder) for inference, evaluation, and training code.
53
+
54
+ ## ☕️ Citation
55
+
56
+ If you find this repository helpful, please consider citing our paper:
57
+
58
+ ```
59
+ @article{yu2023wavecoder,
60
+ title={Wavecoder: Widespread and versatile enhanced instruction tuning with refined data generation},
61
+ author={Yu, Zhaojian and Zhang, Xin and Shang, Ning and Huang, Yangyu and Xu, Can and Zhao, Yishujie and Hu, Wenxiang and Yin, Qiufeng},
62
+ journal={arXiv preprint arXiv:2312.14187},
63
+ year={2023}
64
+ }
65
+ ```
66
+ ## Note
67
+
68
+ 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.