Face­book launches open source Torch­net tool­kit to en­able new AI ex­per­i­ments

OpenSource For You - - FOSS BYTES -

In­tend­ing to take ar­ti­fi­cial in­tel­li­gence (AI) to new lev­els, Face­book has launched Torch­net. The open source soft­ware tool­kit is based on the deep ma­chine learn­ing frame­work Torch, and will foster rapid and col­lab­o­ra­tive de­vel­op­ment of deep learn­ing de­vel­op­ments by the Torch com­mu­nity.

De­buted at the International Con­fer­ence on Ma­chine Learn­ing (ICML) in New York, Torch­net uses AI concepts through a se­ries of boil­er­plate code, key ab­strac­tions and ref­er­ence im­ple­men­ta­tions. The of­fer­ings through the tool­kit can be com­bined or reused sep­a­rately to drive fur­ther de­vel­op­ments among re­searchers in the field of deep ma­chine learn­ing.

“We created Torch­net to give re­searchers clear guide­lines on how to set up their code, and boil­er­plate code that helps them de­velop more quickly. The mod­u­lar Torch­net de­sign makes it easy to test a se­ries of cod­ing vari­ants fo­cused around the data set, the data load­ing process, and the model, as well as op­ti­mi­sa­tion and per­for­mance mea­sures,” the Torch­net team mem­bers, led by Face­book’s Ar­ti­fi­cial Re­search (FAIR) lab, wrote in a blog post.

Torch­net is writ­ten in the multi-par­a­digm pro­gram­ming lan­guage, Lua. It is de­signed to run on any de­vice with a stan­dard x86 chip.

Face­book is not the only tech com­pany that has started view­ing deep learn­ing with keen in­ter­est. Start­ing from Ama­zon and Google to Mi­crosoft, all the lead­ing play­ers are in the race to ex­pand their pres­ence in the world of AI with their of­fer­ings. In fact, Twit­ter is al­ready us­ing the same Torch frame­work to build ma­chine learn­ing soft­ware.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.