このブログでも5月9日に取り上げたGraphCore
引用 Page. 21 (4:18:00頃) ・学習と推論 ( both training and deployment) ・16nm / PCIeカードに2個 ・DNN performance well beyond Volta and TPU; efficient without large batches. ・Program in TensorFlow initially, other o/s frameworks follow; or Poplar for close-to-metal ・Early access cards and appliances end-2017
追記)、2017.7.8(土)
デバイスの名前も公開されていたのですね。
Graphcore's 'Colossus' chip due before end of year
人、めっちゃ、募集している
デバイスの名前も公開されていたのですね。
Graphcore's 'Colossus' chip due before end of year
人、めっちゃ、募集している
Engineering ・Graduate/PhD Software Engineer ・Machine Learning Systems Developer ・Software Infrastructure Engineer ・Software Verification Engineer ・Machine Learning Framework Developer ・Applied Machine Learning Software Engineer ・Host Runtime Software Engineer ・Software Platform Engineer (Debug & Performance Tools) ・Software Platform Engineer (Compiler Specialist) ・Manufacturing Operations Planner / Manager ・Product Engineer ・Software Platform Engineer ・Quality / Reliability Engineer