0000000001183408
AUTHOR
Hieu Hoang
Fast Neural Machine Translation Implementation
This paper describes the submissions to the efficiency track for GPUs at the Workshop for Neural Machine Translation and Generation by members of the University of Edinburgh, Adam Mickiewicz University, Tilde and University of Alicante. We focus on efficient implementation of the recurrent deep-learning model as implemented in Amun, the fast inference engine for neural machine translation. We improve the performance with an efficient mini-batching algorithm, and by fusing the softmax operation with the k-best extraction algorithm. Submissions using Amun were first, second and third fastest in the GPU efficiency track.
Numerical and experimental verification of new method for connecting pipe to flange by cold forming
Abstract In this paper a new method of connecting pipe to flange without welding is presented. This method is a cold forming process that is based on plastic expansion/deformation of the pipe into a modified standard flange by use of a cold forming tool. The method is patented by Quickflange Technology AS and represents a highly feasible alternative to welding. The successful use of the method requires the ability to predict dimensional and stress/strain characteristics of the pipe and flange after the connection process in order to evaluate the connectivity to the adjacent flange as well as the leak tightness. In addition the ability to predict the process force during the connection proce…