Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313442384> ?p ?o ?g. }
Showing items 1 to 99 of
99
with 100 items per page.
- W4313442384 abstract "Deep-learning (DL) compilers such as TVM and TensorRT are increasingly being used to optimize deep neural network (DNN) models to meet performance, resource utilization and other requirements. Bugs in these compilers can result in models whose semantics differ from the original ones, producing incorrect results that corrupt the correctness of downstream applications. However, finding bugs in these compilers is challenging due to their complexity. In this work, we propose a new fuzz testing approach for finding bugs in deep-learning compilers. Our core approach consists of (i) generating diverse yet valid DNN test models that can exercise a large part of the compiler's transformation logic using light-weight operator specifications; (ii) performing gradient-based search to find model inputs that avoid any floating-point exceptional values during model execution, reducing the chance of missed bugs or false alarms; and (iii) using differential testing to identify bugs. We implemented this approach in NNSmith which has found 72 new bugs for TVM, TensorRT, ONNXRuntime, and PyTorch to date. Of these 58 have been confirmed and 51 have been fixed by their respective project maintainers." @default.
- W4313442384 created "2023-01-06" @default.
- W4313442384 creator A5000275316 @default.
- W4313442384 creator A5002058484 @default.
- W4313442384 creator A5002262984 @default.
- W4313442384 creator A5022307899 @default.
- W4313442384 creator A5043546718 @default.
- W4313442384 creator A5069998725 @default.
- W4313442384 creator A5073895064 @default.
- W4313442384 date "2023-01-27" @default.
- W4313442384 modified "2023-10-14" @default.
- W4313442384 title "NNSmith: Generating Diverse and Valid Test Cases for Deep Learning Compilers" @default.
- W4313442384 cites W1480909796 @default.
- W4313442384 cites W1533404702 @default.
- W4313442384 cites W1984762903 @default.
- W4313442384 cites W2002934700 @default.
- W4313442384 cites W2026926213 @default.
- W4313442384 cites W2107147876 @default.
- W4313442384 cites W2121898351 @default.
- W4313442384 cites W2530895108 @default.
- W4313442384 cites W2532737545 @default.
- W4313442384 cites W2759550170 @default.
- W4313442384 cites W2766540688 @default.
- W4313442384 cites W2806377938 @default.
- W4313442384 cites W2808279976 @default.
- W4313442384 cites W2904932877 @default.
- W4313442384 cites W2963150697 @default.
- W4313442384 cites W2997653900 @default.
- W4313442384 cites W3008321987 @default.
- W4313442384 cites W3049735680 @default.
- W4313442384 cites W3104663419 @default.
- W4313442384 cites W3123045479 @default.
- W4313442384 cites W3126096841 @default.
- W4313442384 cites W3152739133 @default.
- W4313442384 cites W3154106427 @default.
- W4313442384 cites W3155524054 @default.
- W4313442384 cites W3177562315 @default.
- W4313442384 cites W3195618452 @default.
- W4313442384 cites W4213287115 @default.
- W4313442384 cites W4221144766 @default.
- W4313442384 cites W4230861591 @default.
- W4313442384 cites W4238083723 @default.
- W4313442384 cites W4284686707 @default.
- W4313442384 cites W4284708930 @default.
- W4313442384 cites W4306823614 @default.
- W4313442384 cites W4308643012 @default.
- W4313442384 doi "https://doi.org/10.1145/3575693.3575707" @default.
- W4313442384 hasPublicationYear "2023" @default.
- W4313442384 type Work @default.
- W4313442384 citedByCount "7" @default.
- W4313442384 countsByYear W43134423842023 @default.
- W4313442384 crossrefType "proceedings-article" @default.
- W4313442384 hasAuthorship W4313442384A5000275316 @default.
- W4313442384 hasAuthorship W4313442384A5002058484 @default.
- W4313442384 hasAuthorship W4313442384A5002262984 @default.
- W4313442384 hasAuthorship W4313442384A5022307899 @default.
- W4313442384 hasAuthorship W4313442384A5043546718 @default.
- W4313442384 hasAuthorship W4313442384A5069998725 @default.
- W4313442384 hasAuthorship W4313442384A5073895064 @default.
- W4313442384 hasBestOaLocation W43134423842 @default.
- W4313442384 hasConcept C108583219 @default.
- W4313442384 hasConcept C119857082 @default.
- W4313442384 hasConcept C154945302 @default.
- W4313442384 hasConcept C169590947 @default.
- W4313442384 hasConcept C199360897 @default.
- W4313442384 hasConcept C41008148 @default.
- W4313442384 hasConcept C50644808 @default.
- W4313442384 hasConcept C55439883 @default.
- W4313442384 hasConceptScore W4313442384C108583219 @default.
- W4313442384 hasConceptScore W4313442384C119857082 @default.
- W4313442384 hasConceptScore W4313442384C154945302 @default.
- W4313442384 hasConceptScore W4313442384C169590947 @default.
- W4313442384 hasConceptScore W4313442384C199360897 @default.
- W4313442384 hasConceptScore W4313442384C41008148 @default.
- W4313442384 hasConceptScore W4313442384C50644808 @default.
- W4313442384 hasConceptScore W4313442384C55439883 @default.
- W4313442384 hasFunder F4320306076 @default.
- W4313442384 hasFunder F4320307757 @default.
- W4313442384 hasFunder F4320307764 @default.
- W4313442384 hasFunder F4320309327 @default.
- W4313442384 hasFunder F4320319290 @default.
- W4313442384 hasLocation W43134423841 @default.
- W4313442384 hasLocation W43134423842 @default.
- W4313442384 hasLocation W43134423843 @default.
- W4313442384 hasOpenAccess W4313442384 @default.
- W4313442384 hasPrimaryLocation W43134423841 @default.
- W4313442384 hasRelatedWork W1497385637 @default.
- W4313442384 hasRelatedWork W1936017528 @default.
- W4313442384 hasRelatedWork W2003848320 @default.
- W4313442384 hasRelatedWork W2044422526 @default.
- W4313442384 hasRelatedWork W2192862863 @default.
- W4313442384 hasRelatedWork W2245390655 @default.
- W4313442384 hasRelatedWork W2369288331 @default.
- W4313442384 hasRelatedWork W2519339279 @default.
- W4313442384 hasRelatedWork W4252501555 @default.
- W4313442384 hasRelatedWork W1602178951 @default.
- W4313442384 isParatext "false" @default.
- W4313442384 isRetracted "false" @default.
- W4313442384 workType "article" @default.