agºÍ¼Ç

À´Ô´£ºÒ©Æ·¹ñÃÜÂë £¬×÷Õߣº £¬£º

°¥Ñ½ £¬½²µ½ËÕÖÝ࣠£¬Äǵط½ÊdzöÁËÃûµÄÎÂÈáϸÄå £¬¹âÊÇÔ°Á־͸ãµÃÈËðò×ÓéKÁË £¬ÕæÏëÌìÌìÎÑÔÚÄÇÀïÆ·ÜøÉ;°¡£µ«ÄãÏþµÃû £¬ËÕÖݵÄÂ¥·ïЧÀÍÒ²Ò»ÑùÓн²¾¿Å¶ £¬ÌرðÊÇ´ºÔËÆÚ¼ä £¬ÈËÀ´ÈËÍù £¬Ìôѡ¥·ï¸üÒª²ÁÁÁÑÛ¾¦¡£½­ËÕËÕÖݵØÇøÂ¥·ï £¬¾¿¾¹¸ãÂÑû£¿ÎÒ¸úÄãÂýÂý°Ú¡£

Â¥·ïÊÇʲôÀ´Í·£¿ËÕÖݵÄÓÐɶ·×ÆçÑù£¿

½²ÂÑʵ»° £¬Â¥·ï˵°×Á˾ÍÊÇÒ»ÖÖ˽ÈËЧÀÍ £¬ÊʺÏÄÇЩÐèÒªËÉ¿ªµÄÀϱíÃÇ¡£ËÕÖݵÄÂ¥·ï°¡ £¬½²µ½Ð§ÀÍ £¬·Ï½²Êǵð¸ÜµÄ£¡ÄãÏþµÃ £¬ËÕÖÝÃÃÖ½ÎÂÈáµÃÊdzöÁËÃûµÄ £¬ÇáÉùϸÓï £¬ÂéÖ±ÈÃÈËÐͼËÖÁË¡£²»¹ýÒ²±ðÖ»¹Ë×ÅÊæ·þ £¬¸ãÂÑû¼ûµÄÄÇÖÖ¾ÍÓеãÂé·³ÂѶ¼µßÁË £¬Ç®»¨ÁË»¹ÐÄ·³¡£

ÐÑÁúµã£º¿ËËÕÖݵÄÂ¥·ïЧÀÍ £¬×îºÃͨ¹ýÕý¹æÇþµÀÕÒ £¬²»È»ÂÑÓö¼Ã» £¬¿ÉÄÜࣻᱻ¿ÓµÃÂÑ»ðÌÌ¡£
Ïà¹ØÍ¼Æ¬

´ºÔËÆÚ¼äÂ¥·ïЧÀÍÓÐɶÐþ»ú£¿

Ïà¹ØÍ¼Æ¬

¶ªÄÇÐÇ £¬´ºÔËÕâ½Ú¹ÇÑÛ¶ù £¬ËÕÖÝÂ¥·ïµÄÐèÇó¿Ï¶¨ÊÇñ²µÃÂѶ¼µø¡£È˶à £¬Ð§ÀÍæ £¬¼Û¸ñÓпÉÄÜࣻáÕÇ¡£ÄãÒªÊǾõµÃ×Ô¼ºÊÇÐÑéÏé­ £¬ÄǾÍÌáǰԤԼ £¬Ã»È»¿ÉÄÜÂѼ·µÃÁ¬¸öµØ¶ù¶¼Ã»¡£¼Ç×Å࣠£¬´ºÔËÆÚ¼äÌôÂ¥·ï £¬±ð¸ãŨˮ £¬ÕÒ¿¿Æ×µÄ²Å²»À˲ÙÐÄÇé¡£

ËÕÖÝÂ¥·ï¼Û¸ñÕ¦Ñù£¿»á²»»á±»¿Ó£¿

½²µã±ðµÄ࣠£¬ËÕÖݵÄÂ¥·ï¼Û¸ñÒ»°ã·Ö²ã´Î £¬»ù´¡Ð§ÀÍËãÂÑ×ÔÖÆ £¬ºÀ»ªÒ»µãµÄ¾ÍÂéÖ±¹óÁË¡£µ«ÄãÒª¼ÇµÃ £¬Ã»ÐÅÄã¿ËËæ±ãÕÒÄÇЩÆßºÚ°ËºÚµÄµØ·½ £¬¸ãÂÑû¼û£¡Õý¹æÂ¥·ïµêÒ»°ã¼Û¸ñ͸Ã÷ £¬¿Ú±®»¹µð¸Ü £¬ÌåÑéˬµ½ÂѶ¼µø¡£ÒªÊÇÅöµ½¹í´òÄãµÄ £¬³¶ÂÑ̸µÄÄÇÖÖ £¬¾Í±ðÀ˲ÙÐÄÇéÁË¡£

ÔõôѡËÕÖÝÂ¥·ï²Å¶¨ÐÄ£¿

ุúÄ㽲࣠£¬ÌôÂ¥·ïЧÀ;ÍÏñÌô¹ðÁÖÃ׷۵ıˮ £¬½²¾¿µÃºÜ¡£µÚÒ» £¬Çé¿öÒª½à¾» £¬¶þÊÇÃÃÖ½Òª¿¿Æ× £¬ÈýÊÇЧÀÍÒªµ½Î»¡£»¹Òª¼Ç×Å࣠£¬±ð±»ÄÇЩ»¨Í·Ñ¼µÄÐû´«¸ãÔÎÁËÍ· £¬Õý¶ù°Ë¾­µÄµê²ÅÊÇÄã¸Ã¿ËµÄµØ·½¡£´ºÔËÆÚ¼äÈ˶à £¬Äã¸ãÂÑûÌáǰԤԼ £¬¿ÉÄÜà£Á¬¸öÃŶ¼½ø²»ÁË¡£


ÄǾ¿¾¹ËÕÖÝÂ¥·ïЧÀÍÖµ²»ÖµµÃÒ»ÊÔ£¿ÂÑÊǵģ¡Ö»ÒªÄãÕçÑ¡¿¿Æ×ÇþµÀ £¬¼Û¸ñ¹«Õý £¬Ð§ÀÍÓÖµð¸Ü £¬·Ö·ÖÖÓÈÃÄãÊæÌ¹µ½ÂѶ¼µø¡£

½­ËÕËÕÖݵØÇøÂ¥·ï, ´ºÔËÂ¥·ï¹¥ÂÔ, ЧÀÍÑ¡Ôñ¼¼ÇÉ, ËÕÖÝÂ¥·ï¼Û¸ñ, Â¥·ïÌåÑéÐĵÃ

Ïà¹ØÍ¼Æ¬

¡¶ÄþµÂ±±ÃŽÖСÏï×ÓÔõô×ß¡·

ÈÕǰ £¬ÓÉXiaoyu MaºÍDavid PattersonÁªºÏÊðÃûµÄÎÄÕ¡¶Challenges and Research Directions for Large Language Model Inference Hardware¡·ÕýʽÐû²¼¡£ÕâÆªÎÄÕ±»Ðû²¼ÒÔºó £¬ÒýÆðÁ˹㷺¹Ø×¢¡£ÎÄÕÂÖÐ £¬×÷ÕßÎ§ÈÆLLMÍÆÀíоƬµÄÌôÕ½ÒÔ¼°½â¾ö¼Æ»® £¬¸ø³öÁ˽¨Òé¡£

¡¶¹ãÖÝÉ£ÄûáËùÂÛ̳¡·

Ïã¸Û½ð¹Ü¾Ö³Æ £¬Ïã¸ÛÀûÂÊÔڿɼûµÄÒ»¶Îʱ¼äÄÚ»òÈÔ´¦Óڽϸßˮƽ £¬ÊÐÃñÔÚ×÷³öÖÃÒµ¡¢°´½Ò»òÆäËû½è´û¾ö׼ʱ £¬Ó¦¼ÌÐøÐ¡ÐÄ¿¼ÂDz¢ÖÎÀíÀûÂÊΣº¦¡£½ð¹Ü¾Ö»á¼ÌÐøÃÜÇмà²ìÊг¡±ä¸ï £¬Î¬³ÖÇ®±Ò¼°½ðÈÚÎȶ¨¡£(Íê)

¡¶Í¬³ÇÐÅÏ¢QQȺ¡·

Ô­ÌâÄ¿£ºÂÞÓÀºÆº°Í£ £¬¼Ö¹úÁúÔÙ·¢Éù£ºËûÈÃÎ÷±´Ñ©ÉϼÓ˪ £¬ÈÃÎÒºÍÔ±¹¤Ôâµ½ÎÞÊýÈèÂî ¸Ä²»ÁË×Ô¼ºµÄÀÏ £¬µ«¿ÉÒÔÂýÂýÈ¥¡°µÇ¡±

ÍøÕ¾µØÍ¼