Matches in SemOpenAlex for { <https://semopenalex.org/work/W4225434596> ?p ?o ?g. }
Showing items 1 to 66 of
66
with 100 items per page.
- W4225434596 abstract "Deep Learning (DL) has shown impressive performance in many mobile applications. Most existing works have focused on reducing the computational and resource overheads of running Deep Neural Networks (DNN) inference on resource-constrained mobile devices. However, the other aspect of DNN operations, i.e. training (forward and backward passes) on smartphone GPUs, has received little attention thus far. To this end, we conduct an initial analysis to examine the feasibility of on-device training on smartphones using mobile GPUs. We first employ the open-source mobile DL framework (MNN) and its OpenCL backend for running compute kernels on GPUs. Next, we observed that training on CPUs is much faster than on GPUs and identified two possible bottlenecks related to this observation: (i) computation and (ii) memory bottlenecks. To solve the computation bottleneck, we optimize the OpenCL backend's kernels, showing 2x improvements (40-70 GFLOPs) over CPUs (15-30 GFLOPs) on the Snapdragon 8 series processors. However, we find that the full DNN training is still much slower on GPUs than on CPUs, indicating that memory bottleneck plays a significant role in the lower performance of GPU over CPU. The data movement takes almost 91% of training time due to the low bandwidth. Lastly, based on the findings and failures during our investigation, we present limitations and practical guidelines for future directions." @default.
- W4225434596 created "2022-05-05" @default.
- W4225434596 creator A5010623957 @default.
- W4225434596 creator A5017587307 @default.
- W4225434596 creator A5021750511 @default.
- W4225434596 creator A5045265829 @default.
- W4225434596 date "2022-02-21" @default.
- W4225434596 modified "2023-09-28" @default.
- W4225434596 title "Enabling On-Device Smartphone GPU based Training: Lessons Learned" @default.
- W4225434596 doi "https://doi.org/10.48550/arxiv.2202.10100" @default.
- W4225434596 hasPublicationYear "2022" @default.
- W4225434596 type Work @default.
- W4225434596 citedByCount "0" @default.
- W4225434596 crossrefType "posted-content" @default.
- W4225434596 hasAuthorship W4225434596A5010623957 @default.
- W4225434596 hasAuthorship W4225434596A5017587307 @default.
- W4225434596 hasAuthorship W4225434596A5021750511 @default.
- W4225434596 hasAuthorship W4225434596A5045265829 @default.
- W4225434596 hasBestOaLocation W42254345961 @default.
- W4225434596 hasConcept C111919701 @default.
- W4225434596 hasConcept C11413529 @default.
- W4225434596 hasConcept C149635348 @default.
- W4225434596 hasConcept C154945302 @default.
- W4225434596 hasConcept C1665295 @default.
- W4225434596 hasConcept C173608175 @default.
- W4225434596 hasConcept C186967261 @default.
- W4225434596 hasConcept C188045654 @default.
- W4225434596 hasConcept C2776214188 @default.
- W4225434596 hasConcept C2780513914 @default.
- W4225434596 hasConcept C3826847 @default.
- W4225434596 hasConcept C41008148 @default.
- W4225434596 hasConcept C45374587 @default.
- W4225434596 hasConcept C516764902 @default.
- W4225434596 hasConcept C60952562 @default.
- W4225434596 hasConceptScore W4225434596C111919701 @default.
- W4225434596 hasConceptScore W4225434596C11413529 @default.
- W4225434596 hasConceptScore W4225434596C149635348 @default.
- W4225434596 hasConceptScore W4225434596C154945302 @default.
- W4225434596 hasConceptScore W4225434596C1665295 @default.
- W4225434596 hasConceptScore W4225434596C173608175 @default.
- W4225434596 hasConceptScore W4225434596C186967261 @default.
- W4225434596 hasConceptScore W4225434596C188045654 @default.
- W4225434596 hasConceptScore W4225434596C2776214188 @default.
- W4225434596 hasConceptScore W4225434596C2780513914 @default.
- W4225434596 hasConceptScore W4225434596C3826847 @default.
- W4225434596 hasConceptScore W4225434596C41008148 @default.
- W4225434596 hasConceptScore W4225434596C45374587 @default.
- W4225434596 hasConceptScore W4225434596C516764902 @default.
- W4225434596 hasConceptScore W4225434596C60952562 @default.
- W4225434596 hasLocation W42254345961 @default.
- W4225434596 hasLocation W42254345962 @default.
- W4225434596 hasOpenAccess W4225434596 @default.
- W4225434596 hasPrimaryLocation W42254345961 @default.
- W4225434596 hasRelatedWork W1472213334 @default.
- W4225434596 hasRelatedWork W1540523210 @default.
- W4225434596 hasRelatedWork W1594846894 @default.
- W4225434596 hasRelatedWork W1905767204 @default.
- W4225434596 hasRelatedWork W2054816578 @default.
- W4225434596 hasRelatedWork W2297325673 @default.
- W4225434596 hasRelatedWork W2887888236 @default.
- W4225434596 hasRelatedWork W3195610113 @default.
- W4225434596 hasRelatedWork W4225434596 @default.
- W4225434596 hasRelatedWork W4229055905 @default.
- W4225434596 isParatext "false" @default.
- W4225434596 isRetracted "false" @default.
- W4225434596 workType "article" @default.