你可以使用光标键移动图表。
Created with Raphaël 2.2.08Apr763231Mar30292827262523181614133Feb22Jan2118631Dec3029282726252418139743229Nov282726252322211918171615141318Oct17151110932130Sep282726252423222120191817161514131211109726Aug252220191816109765432131Jul302928272625232220191817161413121110987654length penaltyxiaotong-workingxiaotong-workingmore methods of accessing tensor entries and creating tensorsclean the codeoverload of +, -, * and /model score with length penaltyoverload of +hypothesis idfunctions for float-typed datarename argumentsbug fixesbug fixes of binary math functionsadd binary math functionsgenerate index of previous states in inferencenew methodsadd Get and Set methods to fetch and save int-type itemsadd CopyValues to rename _CopyValuescoding of predictorgeneration via beam pruninguse Sort instead of TopK if K is larger than the size of the dimension we go along withcode of searcher and predictorcode cleaningadd name of the layerAdd SetDataFixedIntbetter implementation of maskimprove the code of the attention modelcodingclass definitionadd some filessave the encoding output layer for inferenceAdding "name" into XTensor. Now we can search for a node in the network by its "name".better check of inputc++ 11 supportedmore codeimprove the code of maskingdelete amax, asum, etc. delete global variable useBLAS also.litianzhilitianzhidelete amax, asum, etc. delete global variable useBLAS also.delete amax, asum, etc. delete global variable useBLAS also.delete amax, asum, etc. delete global variable useBLAS also.predict the next state given the current stateadd NewTensor() to generate an empty XTensor