Lalacewar yanayin da hayaƙin iskar gas ke fitarwa ya fi bayyane kuma don taimakawa masu binciken don fahimtar gudummawar da hankali ke bayarwa ga canjin yanayi da kuma yin amfani da sabbin sifofin bincike wanda zai rage fitar da hayaki An ɗauka azaman ƙimar aiki mai mahimmanci, ƙungiyar masu binciken AI na ƙasa da ƙasa da masana kimiyyar bayanai sun haɗa kai don tsara software wanda zai iya kimanta sawun ƙarancin ayyukan IT.
CodeCarbon bude babbar hanyar software an tsara shi don taimaka wa kamfanoni saka idanu kan sawun ƙarancin AI.
Comet, mai ba da mafita na MLOps, ya yi aiki tare da ƙungiyar AI da kamfanonin kimiyyar bayanai daga ko'ina cikin duniya: MILA, ɗakin binciken AI wanda Yoshua Bengio ya jagoranta a Montreal, BCG GAMMA, sashen nazarin da kimiyyar bayanai daga Boston Rukuni mai ba da shawara da Kwalejin Haverford a Pennsylvania, don ƙirƙirar software ta buɗewa.
Game da CodeCarbon
CodeCarbon software ne Python tushen que zai ba masu shirye-shirye damar sanya lambar su mafi inganci kuma rage adadin CO2 da aka samar don amfani da kayan sarrafa kwamfuta kuma zai motsa su suyi hakan.
Manhajar ba wai kawai kimanta adadin CO2 da aka samar ba don amfani da kayan IT, hakanan yana baiwa masu kirkira shawara kan yadda zasu rage fitar da hayaki zaɓar kayan aikin girgijen ku a yankuna masu amfani da ƙananan hanyoyin samar da makamashi.
Yoshua Bengio, wanda ya kafa MILA kuma ya ci kyautar Turing, ya ce:
“AI fasaha ce mai ƙarfi kuma mai ƙarfi ga mai kyau, amma yana da mahimmanci a san irin tasirin tasirin muhalli. Aikin CodeCarbon yana da manufar daidai don cimma wannan burin kuma ina fata zai sa jama'ar AI su kirga, su bayyana tare da rage sawun carbon ɗin su. ”
Sylvain Duranton, Manajan Darakta kuma Babban Aboki a kungiyar Boston Consulting Group (BCG) da kuma Darakta na Duniya a BCG GAMMA, ya ce:
“Dangane da tarihin kwanan nan, amfani da IT gaba ɗaya, da AI musamman, za su ci gaba da haɓaka ƙwarai a duk duniya. A cikin wannan mahallin, CodeCarbon na iya taimakawa ƙungiyoyi don tabbatar da cewa sawun kawunansu na karɓa yana ƙaruwa sosai kamar yadda ya kamata ”.
A cikin zurfin ilimin bincike mai zurfin ilmantarwa, ana samun ci gaba a cikin ilimin kere kere ta hanyar ƙirƙirar manyan samfuran, tattara manyan bayanai, da kuma amfani da ikon sarrafa kwamfuta mafi girma.
Horar da algorithm mai ilmantarwa mai ƙarfi na iya buƙatar amfani da kwamfutoci da yawa a cikin kwanaki ko makonni.
Don gine-gine kamar VGG, BERT, GPT-2 da GPT-3, waɗanda ke da miliyoyin abubuwan daidaitawa kuma ana horar dasu akan GPU da yawa na makonni da yawa, wannan na iya zama banbancin kilo kilo dari na CO-eq.
OpenAI's GPT-2 da aka ƙaddamar a cikin 2019 ya dogara ne da sigogi biliyan 1.5, yayin da magajinsa GPT-3 aka ƙaddamar a shekarar da ta gabata, wanda sigogin biliyan 175 ya sa ya ninka wanda ya gabata sau 100. Yayin da manyan samfura ke ci gaba a fagen, yawan kuzarin da ake amfani da su don horar da su zai ƙaru.
CodeCarbon yana da tsarin tsarin bin diddigin adadin makamashin da aka yi amfani da shi ta hanyar jagorantar masu samar da lissafi na girgije da kuma keɓaɓɓun cibiyoyin bayanai.
Bayan haka, tsarin yana amfani da bayanai daga kafofin jama'a don kimanta ƙimar CO2 da aka samu, Tabbatar da ƙididdigar cibiyar sadarwar lantarki wanda aka haɗa kayan aikin.
Maɓallin ya kiyasta CO2 da aka samar don kowane gwaji ta amfani da takamaiman ƙirar AI, adana bayanan fitarwa don ayyukan da kuma ga ƙungiyar duka.
Tunanin shine CodeCarbon zai taimaka wa kamfanonin IT da AI don iyakance sawun ƙarancin carbon ɗin su yayin da suke girma. CodeCarbon zai samar da dashboard wanda zai baiwa kamfanoni damar ganin yawan fitowar hayakin da ake fitarwa ta hanyar horar da tsarin koyon injinansu.
Toarfin bin diddigin hayaƙin CO2 yana wakiltar ci gaba mai mahimmanci a cikin ikon masu haɓaka don amfani da albarkatun makamashi cikin hikima sabili da haka rage tasirin aikinsu cikin mawuyacin yanayi.
Source: https://www.comet.ml/