GitHub emepeela mmepe n'iji mmụta igwe maka ọchụchọ na nyocha koodu

GitHub webatara ọrụ ahụ CodeSearchNet, nke akwadola ụdị mmụta igwe na nhazi data dị mkpa maka ntule, nhazi na nyocha koodu n'asụsụ mmemme dị iche iche. CodeSearchNet, yiri Ihe ntanetị, gụnyere nnukwu nchịkọta koodu snippets nwere nkọwa ndị na-emezi ihe koodu ahụ na-eme. Ngwa maka ụdị ọzụzụ na ihe atụ nke iji CodeSearchNet ka edere na Python site na iji usoro Tensorflow na kesara site n'okpuru ikike MIT.

Mgbe ị na-eke CodeSearchNet, ejiri teknụzụ nyocha ederede asụsụ eke, na-eme ka sistemu mmụta igwe nwee ike iburu n'uche ọ bụghị naanị njirimara syntactic, kamakwa nkọwa nke omume ndị koodu ahụ mere. Sistemụ GitHub etinyere ya N'ime nnwale na ịhazi koodu nyocha site na iji ajụjụ na eke asụsụ (dịka ọmụmaatụ, mgbe ị na-arịọ "ịhazi ndepụta nke eriri", koodu na mmejuputa algọridim kwekọrọ na-egosipụta).

Nchịkọta data echere na-agụnye ihe karịrị njikọ nkọwa koodu nde abụọ, akwadoro dabere na ederede isi mmalite nke ọba akwụkwọ mepere emepe dị ugbu a. Koodu ahụ na-ekpuchi ederede isi mmalite zuru oke nke ọrụ ma ọ bụ ụzọ onye ọ bụla, nkọwa ahụ na-akọwakwa omume ndị ọrụ ahụ rụrụ (a na-enye nkọwa zuru ezu). Ugbu a, a na-akwado ntọala data maka Python, JavaScript, Ruby, Go, Java na PHP. Enyere ihe atụ nke iji datasets a chọrọ maka ịzụ ụdị netwọkụ akwara dị iche iche, gụnyere Akpa akwara-Okwu, RNN, Nlebara anya onwe onye (BERT) na Ngwakọ 1D-CNN+Nlebara Onwe Onye.

Iji zụlite usoro ịchọ asụsụ okike, akwadokwala ọtụtụ CodeSearchNet Challenge, gụnyere
99 nkịtị ajụjụ nwere ihe dị ka puku nkọwa ndị ọkachamara 4, na-akọwa njikọ koodu nwere ike na CodeSearchNet Corpus dataset, na-ekpuchi ihe dịka ụzọ na ọrụ nde isii (6).setịpụrụ nha ihe dịka 20 GB). Ihe ịma aka CodeSearchNet nwere ike ije ozi dị ka ihe nrịbama maka nyocha ịdị irè nke ụfọdụ ụzọ maka ịchọ koodu asụsụ eke. Iji ngwá ọrụ KubeFlow kwadebere
ihe atụ koodu search engine.

isi: opennet.ru

Tinye a comment