5M ݳ OpenAI Code Embedding ģ OASIS ()
2024-12-02 21:52:45 Դ IT֮
ղ(0)
ɨά
ѺȦ
QQռ ɨһɨ
ɨά
ѺȦ
Kwaipilot ŶʽԴ OASISOptimized Augmentation Strategy for Improved code SearchCode Embedding ģ͡Kwaipilot ϷƼѧ Arise ʵҽкУڴʵͻԽչʹ 5M Tokens ݴ OpenAIڶ Code Search benchmark Խˮƽ
ʲôǴ (Code Embedding)?
ŴģijԽԽЧĴϵͳЧʡͳĹؼƥ䷽ѾִCode EmbeddingͨƬתΪʾʹܹ壬ӶʵָܵĴCode Embedding ģڴֿ⼶ʴ𣬴벹ȫȳзӲɻȱá
ͼ 1. [1]
OASIS ЩڿƼ
OASIS ѵֻʹ 5M Tokens ݣԶ SOTA ģ͵ѵͨϳ봴ƵʧԼ͵ijɱڶԼϳԽе SOTA ģ͡
˵OASIS ģͲ˶¼
1.ֿ⼶ͳעƬΣȱٴϢϷƼѧ Arise ʵҼOASIS ˲ֿ⼶ijͨùϵṹģܹõڸе壬ڹʹú docstringϢԡ
2.OASIS-instruct ݺϳ㷨ǿרǿܹԶɸѵЩȻԵĶӦϵʹģܹѧϰϸ졣
3.ںʽʧԵؽĿŻӦģѵȷģڱִͳ Code Embedding ģ£ȷʶϸ죬ŻĿĽǶµѵάȡ
OASIS жǿ
ȨĴУOASIS չֳĿܣ
OASIS ѵвԼṩѵ£ CSNCoSQAAdvTest ⼯ϣԽƽȷͬߴ Code Embedding ģͣ˵ 5M ѵͬĴݼȫ泬Խ OpenAI-Ada-002ʹõ Codesage-Large £ CoSQA ݼϳ 7.8%ģͳߴΪ֮һƽɹȫ泬Խ CodeFuse-CGE-Smallݼԡ
CodeSearchNet (CSN) ݼ
ݼ 200 -ĵԣ PythonJavaJavaScript ԡҪԴںĴ뼰ӦĵַdocstringʹȻԵ docstring Ϊѯquery
CoSQA ݼ
CoSQA ݼһ 20,000 + ȻԲѯʹıעݼÿݶԶ 3 ע߽бעCoSQA ݼеIJѯԴʵӳ˿ճľƬӲͬݼоɸѡȷԺ
AdvTest ݼ
AdvTest ݼרڲԴѶȡͨ淶ͱսԣԭʼй˳ݼ 2 IJּģڴܡ
OASIS ģͽʹ 5M ѵݣ 1.3B IJ CSNCoSQAAdvTest ݼ˳ SOTA ģܱͣԶ OpenAI-Embedding-Ada-002
OASIS ЩӦó
OASIS Ӧó㷺
ܴ棬OASIS ȷߵIJѯͼӺоλʵƬΡʱģܴӺоλʵƬΣƼ뵱ǰĿջƥʾ
ͼ 2. IDE
ڴƼOASIS ͻ˴ͳ벹ȫߵľޡڶԴ⣬ģԤߵıͼƼ API кĹʵַ Kwaipilot 벹ȫ˽뷽еIJȫ
Դ븴÷OASIS ܹȷʶƵʵϸڲͬĴƬΡһ Code Review ģʶеĿܴ
ͼ 3. Kwaipilot
ڴ⸨棬ģӦ Kwaipilot RepoChat ܣģܹԶȡϵͳеĹؼȷĹ͵ùϵͼף߿Ϥֲֿ⣬ŶӵĴɱ
ͬʱΪƶķչOASIS Դ߿ֱͨ Hugging Face ʹøģͣҲԽнһĸĽͶơ
עͬ˳ƾths518ȡ
0
+1
ղ(0)
ɨά
ѺȦ
ɨά
ѺȦ