The new European HPC facility for Fusion is in operation since July 2016. It replaces, for European fusion researchers, the Helios supercomputer installed in Japan in the context of the Broader Approach agreement. The supercomputer is hosted at CINECA and it is a fraction of the MARCONI system. Thanks to a customized technical project done by ENEA, in a joint development agreement with CINECA, the European community of fusion modelling can exploit the latest available CPU technologies, following the CINECA HPC roadmap towards 50 PFlops planned for 2019. The MARCONI Fusion fraction is being delivered in two phases: the first one, 1 PFlops of CPU multi-core architecture based on the Intel Broadwell processors, is already in operation since July 2016, and the second one, 5 PFlops of the same architecture based on the INTEL Skylake processors, will be deployed in July 2017. Furthermore the project includes 1 PFlops of the third generation of Intel Xeon Phi many-core architecture (Knights Landing generation). Within this framework, ENEA/CINECA provides, in addition, the operation support of the Gateway infrastructure of EUROFusion Work-Package Code Development. A new Gateway HPC system is in operation at CINECA since Jan. 2017 thanks to the data migration and software porting activities carried out by ENEA/CINECA team together with the Core Programming Team of the Infrastructure and Support Activity work package from EUROfusion. The new Gateway infrastructure is tightly coupled with the MARCONI Fusion fraction, sharing the same 100 Gbps low-latency network based on the Intel OmniPath technology. The paper describes the technical details and the performances of MARCONI, one of the largest HPC OmniPath based infrastructure.

Data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Pythonbased natural language augmentation framework which supports the creation of both transformations (modifications to the data) and filters (data splits according to specific features). We describe the framework and an initial set of 117 transformations and 23 filters for a variety of natural language tasks. We demonstrate the efficacy of NL-Augmenter by using several of its tranformations to analyze the robustness of popular natural language models. The infrastructure, datacards and robutstness analysis results are available publicly on the NL-Augmenter repository (https://github. com/GEM-benchmark/NL-Augmenter).

No abstract

No abstract

No abstract