Παρακαλώ χρησιμοποιήστε αυτό το αναγνωριστικό για να παραπέμψετε ή να δημιουργήσετε σύνδεσμο προς αυτό το τεκμήριο:
https://hdl.handle.net/20.500.14279/33101
Τίτλος: | XFeatur: Hardware Feature Extraction for DNN Auto-tuning | Συγγραφείς: | Acosta, Jorge Sierra Diavastos, Andreas Gonzalez, Antonio |
Major Field of Science: | Engineering and Technology | Field Category: | Computer and Information Sciences | Ημερομηνία Έκδοσης: | 2022 | Πηγή: | IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS), 2022, 22-24 May | Conference: | IEEE International Symposium on Performance Analysis of Systems and Software, ISPASS 2022 | Περίληψη: | In this work, we extend the auto-tuning process of the state-of-the-art TVM framework with XFeatur; a tool that extracts new meaningful hardware-related features that improve the quality of the representation of the search space and consequently improve the accuracy of its prediction algorithm. These new features provide information about the amount of thread-level parallelism, shared memory usage, register usage, dynamic instruction count and memory access dependencies. Optimizing ResNet-18 with the proposed features improves the quality of the search space representation by 63% on average and a maximum of 2× for certain tasks, while it reduces the tuning time by 9% (approximately 1.1 hours) and produces configurations that have equal or better performance (up to 92.7%) than the baseline. | URI: | https://hdl.handle.net/20.500.14279/33101 | ISBN: | 9781665459549 | DOI: | 10.1109/ISPASS55109.2022.00013 | Rights: | © IEEE | Type: | Conference Papers | Affiliation: | Universitat Politècnica de Catalunya |
Εμφανίζεται στις συλλογές: | Δημοσιεύσεις σε συνέδρια /Conference papers or poster or presentation |
CORE Recommender
Όλα τα τεκμήρια του δικτυακού τόπου προστατεύονται από πνευματικά δικαιώματα