DC25_SLIDES_Track14_FoundationalModelApproachForSIPI_Kashyap.pdf

1、Information Classification:GeneralWelcome to ConferenceJanuary 2830,2025Santa Clara Convention Center1ExpoJanuary 2930,2025Information Classification:GeneralFoundational Model Approach for SI/PI Analysis Using Large Language Model Techniques Priyank Kashyap,(Hewlett-Packard Enterprise)Priyank Kashya

2、p(Hewlett-Packard Enterprise),Yuejiang Wen(Hewlett-Packard Enterprise),YongJin Choi(Hewlett-Packard Enterprise),Chris Cheng(Hewlett-Packard Enterprise)2Information Classification:General Self-attention enables the model to learn the impact of different parts of sequence*o Multiple self-attention mod

3、ules allow for paying attention to different parts of a sequenceo Faster than an LSTM Forms the backbone for large language models Has enabled lots of work in SI/PI analysisTransformers Can Do It All!3Fig.A transformer encoder with its individual components.*A.Vaswani,N.Shazeer,N.Parmar,et al.,“Atte

4、ntion is all you need,”Advances in Neural Information Processing Systems,vol.30,2017.Information Classification:General Routing designs Decision transformers 1 Modeling S-parameters -Encoder+Decoder models-2-3 Thermal analysis 4Transformers Can Do It All!41 M.Kim et al.,“Neural Language Model Enable

5、s Extremely Fast and Robust Routing on Interposer,”2021 DesignCon2 H.Park et al.,High-speed Channel Simulator using Neural Language Models,2024 IEEE International Symposium on Electromagnetic Compatibility,Signal&Power Integrity(EMC+SIPI),Phoenix,AZ,USA,2024,pp.11-16,doi:10.1109/EMCSIPI49824.2024.10

6、705639.3 H.Park et al.,High-Speed Channel Transformer:A Scalable Transformer Network-Based Signal Integrity Simulator,in IEEE Transactions on Electromagnetic Compatibility,doi:10.1109/TEMC.2024.3442232.,4 Lu J,Tan S X,Thermal Map Dataset for Commercial Multi/Many Core CPU/GPU/TPU,Proceedings of the