Matches in SemOpenAlex for { <https://semopenalex.org/work/W4306294911> ?p ?o ?g. }
Showing items 1 to 69 of
69
with 100 items per page.
- W4306294911 abstract "We revisit the performance of the classic gradual magnitude pruning (GMP) baseline for large language models, focusing on the classic BERT benchmark on various popular tasks. Despite existing evidence in the literature that GMP performs poorly, we show that a simple and general variant, which we call GMP*, can match and sometimes outperform more complex state-of-the-art methods. Our results provide a simple yet strong baseline for future work, highlight the importance of parameter tuning for baselines, and even improve the performance of the state-of-the-art second-order pruning method in this setting." @default.
- W4306294911 created "2022-10-15" @default.
- W4306294911 creator A5074310887 @default.
- W4306294911 creator A5083822059 @default.
- W4306294911 date "2022-10-12" @default.
- W4306294911 modified "2023-09-29" @default.
- W4306294911 title "GMP*: Well-Tuned Gradual Magnitude Pruning Can Outperform Most BERT-Pruning Methods" @default.
- W4306294911 doi "https://doi.org/10.48550/arxiv.2210.06384" @default.
- W4306294911 hasPublicationYear "2022" @default.
- W4306294911 type Work @default.
- W4306294911 citedByCount "0" @default.
- W4306294911 crossrefType "posted-content" @default.
- W4306294911 hasAuthorship W4306294911A5074310887 @default.
- W4306294911 hasAuthorship W4306294911A5083822059 @default.
- W4306294911 hasBestOaLocation W43062949111 @default.
- W4306294911 hasConcept C108010975 @default.
- W4306294911 hasConcept C111368507 @default.
- W4306294911 hasConcept C111472728 @default.
- W4306294911 hasConcept C119857082 @default.
- W4306294911 hasConcept C121332964 @default.
- W4306294911 hasConcept C126691448 @default.
- W4306294911 hasConcept C12725497 @default.
- W4306294911 hasConcept C127313418 @default.
- W4306294911 hasConcept C1276947 @default.
- W4306294911 hasConcept C137293760 @default.
- W4306294911 hasConcept C138885662 @default.
- W4306294911 hasConcept C154945302 @default.
- W4306294911 hasConcept C185798385 @default.
- W4306294911 hasConcept C205649164 @default.
- W4306294911 hasConcept C2780586882 @default.
- W4306294911 hasConcept C41008148 @default.
- W4306294911 hasConcept C58640448 @default.
- W4306294911 hasConcept C6557445 @default.
- W4306294911 hasConcept C86803240 @default.
- W4306294911 hasConceptScore W4306294911C108010975 @default.
- W4306294911 hasConceptScore W4306294911C111368507 @default.
- W4306294911 hasConceptScore W4306294911C111472728 @default.
- W4306294911 hasConceptScore W4306294911C119857082 @default.
- W4306294911 hasConceptScore W4306294911C121332964 @default.
- W4306294911 hasConceptScore W4306294911C126691448 @default.
- W4306294911 hasConceptScore W4306294911C12725497 @default.
- W4306294911 hasConceptScore W4306294911C127313418 @default.
- W4306294911 hasConceptScore W4306294911C1276947 @default.
- W4306294911 hasConceptScore W4306294911C137293760 @default.
- W4306294911 hasConceptScore W4306294911C138885662 @default.
- W4306294911 hasConceptScore W4306294911C154945302 @default.
- W4306294911 hasConceptScore W4306294911C185798385 @default.
- W4306294911 hasConceptScore W4306294911C205649164 @default.
- W4306294911 hasConceptScore W4306294911C2780586882 @default.
- W4306294911 hasConceptScore W4306294911C41008148 @default.
- W4306294911 hasConceptScore W4306294911C58640448 @default.
- W4306294911 hasConceptScore W4306294911C6557445 @default.
- W4306294911 hasConceptScore W4306294911C86803240 @default.
- W4306294911 hasLocation W43062949111 @default.
- W4306294911 hasOpenAccess W4306294911 @default.
- W4306294911 hasPrimaryLocation W43062949111 @default.
- W4306294911 hasRelatedWork W2583402933 @default.
- W4306294911 hasRelatedWork W2611669587 @default.
- W4306294911 hasRelatedWork W2913647591 @default.
- W4306294911 hasRelatedWork W2951714314 @default.
- W4306294911 hasRelatedWork W2963941635 @default.
- W4306294911 hasRelatedWork W3199608561 @default.
- W4306294911 hasRelatedWork W4205675155 @default.
- W4306294911 hasRelatedWork W4288635965 @default.
- W4306294911 hasRelatedWork W4293763776 @default.
- W4306294911 hasRelatedWork W4306294911 @default.
- W4306294911 isParatext "false" @default.
- W4306294911 isRetracted "false" @default.
- W4306294911 workType "article" @default.