æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ããä»åã¯ãLLM (Large Language Models) ãæ´»ç¨ãã¦ãæ©æ¢°å¦ç¿ã¢ãã«ã«å¿ è¦ãªãã¼ã¿ã®ã¢ããã¼ã·ã§ã³ä½æ¥ãå¹çåããåãçµã¿ã«ã¤ãã¦ç´¹ä»ãã¾ãã ãªããã¢ããã¼ã·ã§ã³ã«ãããLLMã®å©ç¨ã«é¢ãã¦ã¯ãã¯ã©ã¦ããã³ãã¼å社åâ¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ããåå㯠NVIDIA Triton Inference Server ã®æ§è½ãæ¤è¨¼ãã話ãæ¸ãã¾ãããä»åã¯ãã®ç¶ç·¨ã¨ãªãã¾ãã tech.layerx.co.jp ååã®è¨äºä»¥éãç¶ç¶ãã¦Triton Inference Serverã®æ¤è¨¼ãéããçµæãæ¨è«ãµã¼ãã®æ§è½ãå¤§å¹ ã«æ¹â¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ããä»åã¯æ©æ¢°å¦ç¿ã¢ãã«ã®æ¨è«ãµã¼ãã¨ã㦠NVIDIA Triton Inference Server ã®æ§è½ãæ¤è¨¼ãã話ã§ãã (追è¨) ç¶ç·¨ãæ¸ãã¾ãã tech.layerx.co.jp èæ¯ ãã¯ã©ã¯ã§ã¯è«æ±æ¸OCRãã¯ããã¨ããæ©æ¢°å¦ç¿ã¢ãã«ãéçºãã¦ãã¾ããâ¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ãã ãã®è¨äºã¯ DEIM2024 (第16åãã¼ã¿å·¥å¦ã¨æ å ±ããã¸ã¡ã³ãã«é¢ãããã©ã¼ã©ã ) ã®åå ã¬ãã¼ãã¨ãªãã¾ããLayerXã¯æ¨å¹´ã«å¼ãç¶ããã©ããã¹ãã³ãµã¼ã¨ãã¦åè³ããã¦ããã ãã¾ããã tech.layerx.co.jp ä»å¹´ã¯ 2æ28æ¥(â¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ãã ãã®è¨äºã¯LayerXããã¯ã¢ãã«ã¬14æ¥ç®ã®è¨äºã§ããåå㯠@shnjtk ã«ãã ã¹ãã¼ãªã¼ãã¤ã³ãã§ã¯ãªãã¢ã¦ãã«ã ã§éçºé度ã測ã ã§ããã次å㯠osuke ãããæ å½ãã¾ãã ä»åã¯ãã¯ã©ã¯ã®å¸³ç¥¨ç»åã使ã£ã¦DALL-Eã®dVAE â¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ããæ¬è¨äºã§ã¯ãLayoutLMv3*1ã¨ããã¢ãã«ããã¯ã©ã¯ã§åãæ±ã£ã¦ãã帳票ã§äºåå¦ç¿ãè¡ããããããã¡ã¤ã³ãã¥ã¼ãã³ã°ãã¦é ç®æ¨å®ã¿ã¹ã¯ã«åãçµãã§ãã話ããç´¹ä»ãã¾ãã èæ¯ LayerXã§æä¾ãã¦ãããã¯ã©ã¯ã§ã¯å¸³ç¥¨ãã¢â¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ãã å¤ã§ããã7æã¯LayerXã¨ã³ã¸ãã¢ããã°ãæ´»çºã«ãããæé ã§ãã æ¨å¹´ãã¯ã©ã¯OCRã®æ©æ¢°å¦ç¿ã¢ãã«ã®æ¤è¨¼ããæ¬çªæå ¥ã¾ã§ã®åãçµã¿ã«ã¤ãã¦è¨äºãæ¸ãã¾ããã tech.layerx.co.jp ãã®å¾ãéç¨ããä¸ã§æ°ããªèª²é¡ãçã¾â¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ãã ãã®è¨äºã¯2023å¹´6æ6æ¥(ç«) ~ 6æ9æ¥(é)ã«çæ¬ã§éå¬ãããJSAI2023 (第37å 人工ç¥è½å¦ä¼å ¨å½å¤§ä¼) ã®åå ã¬ãã¼ãã¨ãªãã¾ãã LayerXã¨ãã¦ä»å¹´ã¯DEIMãNLPã«ç¶ã3åç®ã®å¦ä¼åå ã¨ãªãã¾ããã ä»åJSAIã«åãã¦ãã©ãâ¦
æ©æ¢°å¦ç¿ã¨ã³ã¸ãã¢ã®åç°ã§ããä»å¹´ã®3æã«LayerXã«å ¥ç¤¾ã主ã«æ©æ¢°å¦ç¿ãæ´»ç¨ããOCRã®éçºããã¦ãã¾ããããã¾ã§ã«ã¼ã«ãã¼ã¹ã§å®è£ ããã¦ããå·¥ç¨ãä¸é¨æ©æ¢°å¦ç¿ãã¼ã¹ã®å®è£ ã«ç½®ãæãã¤ã¤ããå®éã«æ¬çªã§ã稼åãã¦ãã¾ãããã®è¨äºã§ã¯ã©ã®ããã«æ¤â¦