A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Filters
PET: Optimizing Tensor Programs with Partially Equivalent Transformations and Automated Corrections
2021
USENIX Symposium on Operating Systems Design and Implementation
Preserved Fulltext
We propose PET, the first DNN framework that optimizes tensor programs with partially equivalent transformations and automated corrections. ...
optimized programs by combining fully and partially equivalent optimizations at the tensor, operator, and graph levels. ...
This work is partially supported by National Natural Science Foundation of China (U20A20226, 62072262) and Beijing Natural Science Foundation (4202031). ...
dblp:conf/osdi/WangZGMTZLRCJ21
fatcat:4sf4vbr4avfffhwfdedpidq7oa
DeepEvolution: A Search-Based Testing Approach for Deep Neural Networks
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
One common denominator of these testing techniques is the automated generation of test cases, e.g., new inputs transformed from the original training data with the aim to optimize some test adequacy criteria ...
So far, the effectiveness of these approaches has been hindered by their reliance on random fuzzing or transformations that do not always produce test cases with a good diversity. ...
One common denominator of these testing techniques is the automated generation of test cases, e.g., new inputs transformed from the original training data with the aim to optimize some test adequacy criteria ...
arXiv:1909.02563v1
fatcat:afz7fvaz3zbera235glfvjueeq
Sommelier: Curating DNN Models for the Masses
2022
Proceedings of the 2022 International Conference on Management of Data
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
In this paper, we present Sommelier, an indexing and query system above typical DNN model repositories to interface directly with inference serving or other use cases. ...
Motivated by manual iterative model search processes and typical model design strategies that generate model variants or models with common segments, Sommelier organizes DNN models based on their semantic ...
CNS-1815115 and 2112562, and a Google Faculty Research Award. ...
doi:10.1145/3514221.3526173
fatcat:evlxt33ogbdevbeswanfswtvj4
Special Session: Towards an Agile Design Methodology for Efficient, Reliable, and Secure ML Systems
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
and secure ML systems based on user-defined constraints and objectives. ...
This article summarizes the main challenges in agile development of efficient, reliable and secure ML systems, and then presents an outline of an agile design methodology to generate efficient, reliable ...
Graph Optimizations: Graph-level optimizations help improve computational and memory efficiency by applying various transformations. ...
arXiv:2204.09514v1
fatcat:ho7auszvmferrn36evs7oqdpt4
NeuRI: Diversifying DNN Generation via Inductive Rule Inference
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2023 pre-print arXiv:2302.02261v1 fatcat:2bh5zpqq2fg57k5d5htzowxyoyOther Versions
As such, the recent wave of research has been studying the automated synthesis of test-cases (i.e., DNN models and their inputs) for fuzzing DL systems. ...
NeuRI finds 100 new bugs for PyTorch and TensorFlow in four months, with 81 already fixed or confirmed. ...
Consequently, it is crucial to harness the correctness of DL systems via extensive and automated testing. ...
arXiv:2302.02261v3
fatcat:l3wkyicuuzccbdcbmgj2ndg66i
Refactoring Neural Networks for Verification
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Unlike with traditional code refactoring, DNN refactoring does not guarantee functional equivalence of the two networks, but rather it aims to preserve the accuracy of the original network while producing ...
A DNN refactoring defines (a) the transformation of the DNN's architecture, i.e., the number and size of its layers, and (b) the distillation of the learned relationships between the input features and ...
First, the configuration language provides a simple and unified way to specify architectural transformations, and R4V removes potential errors associated with the transformation itself by automating this ...
arXiv:1908.08026v1
fatcat:j4bgxitgcjdupimxmzxsh4appq
TAP: Accelerating Large-Scale DNN Training Through Tensor Automatic Parallelisation
[article]
2023
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2023; you can also visit the original URL.
The file type is application/pdf.
Experiments show that TAP is 20×- 160× faster than the state-of-the-art automatic parallelism framework, and the performance of its discovered schedules is competitive with the expert-engineered ones. ...
In this work, we present a model parallelism framework TAP that automatically searches for the best data and tensor parallel schedules. ...
However, Y 0 is just a partial result, and it needs to be added with Y 1 from device 1 to get Y . As such, an AllReduceSum communication is required to sum them up for mathematical equivalence. ...
arXiv:2302.00247v1
fatcat:edwlqi7pqng2dd4vjxvwih2lse
Joint Program and Layout Transformations to Enable Convolutional Operators on Specialized Hardware Based on Constraint Programming
2022
ACM Transactions on Architecture and Code Optimization (TACO)
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
The success of Deep Artificial Neural Networks (DNNs) in many domains created a rich body of research concerned with hardware accelerators for compute-intensive DNN operators. ...
However, implementing such operators efficiently with complex hardware intrinsics such as matrix multiply is a task not yet automated gracefully. ...
The produced results also motivate research into the interplay of data layouts, their transformation, and how this interacts with the overall DNN and accelerator architecture. ...
doi:10.1145/3487922
fatcat:gnuvco7rffcdzcuirotonjnssi
HybridDNN: A Framework for High-Performance Hybrid DNN Accelerator Design and Implementation
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
To speedup Deep Neural Networks (DNN) accelerator design and enable effective implementation, we propose HybridDNN, a framework for building high-performance hybrid DNN accelerators and delivering FPGA-based ...
This demonstrates that HybridDNN is flexible and scalable and can target both cloud and embedded hardware platforms with vastly different resource constraints. ...
ACKNOWLEDGMENTS This work is supported in part by the IBM-Illinois Center for Cognitive Computing Systems Research (C3SR) and XMotors.ai. ...
arXiv:2004.03804v1
fatcat:2r7ymftbordw5odrfndowsuxg4
Towards Proof Synthesis Guided by Neural Machine Translation for Intuitionistic Propositional Logic
[article]
2017
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
We implement the whole framework and empirically observe that a generated proof term is close to a correct proof in terms of the tree edit distance of AST. ...
Inspired by the recent evolution of deep neural networks (DNNs) in machine learning, we explore their application to PL-related topics. ...
We also appreciate Akihiro Yamamoto; the discussion with him leads to the evaluation metrics used in this paper. This paper is partially supported by JST PRESTO Grant Number JP-MJPR15E5, Japan. ...
arXiv:1706.06462v1
fatcat:7xho4a42dvfhrn4tu7tridyy7a
Joint Program and Layout Transformations to enable Convolutional Operators on Specialized Hardware based on Constraint Programming
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Other Versions
The success of Deep Artificial Neural Networks (DNNs) in many domains created a rich body of research concerned with hardware accelerators for compute-intensive DNN operators. ...
However, implementing such operators efficiently with complex hardware intrinsics such as matrix multiply is a task not yet automated gracefully. ...
This optimization can reduce the number of nodes and edges in a graph significantly, without loosing correctness of representation. • Nodes with labels with only an outgoing self-edge, or no outgoing edges ...
arXiv:2104.04731v4
fatcat:g4uzwaivtjhajjwaimdct7y5wu
Automated Deep Learning: Neural Architecture Search Is Not the End
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
It requires grappling with problem formulation and context understanding, data engineering, model development, deployment, continuous monitoring and maintenance, and so on. ...
Consequently, in response to these issues, a new field has emerged over the last few years: automated deep learning (AutoDL). ...
Acknowledgments: XD and DJK acknowledge financial support secured by BG, which funded their participation in this study and the ongoing "Automated and Autonomous Machine Learning" project as part of the ...
arXiv:2112.09245v3
fatcat:dujfh7pzmzbrtdyoshkl4kpbsm
Mistify: Automating DNN Model Porting for On-Device Inference at the Edge
2021
Symposium on Networked Systems Design and Implementation
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
This often necessitates fitting a model originally designed and trained in the cloud to edge devices with a range of hardware capabilities, which so far has relied on time-consuming manual effort. ...
Extensive evaluation shows that Mistify reduces the DNN porting time needed by over 10⇥ to cater to a wide spectrum of edge deployment scenarios, incurring orders of magnitude less manual effort. ...
Acknowledgments We thank the anonymous reviewers and our shepherd, Eric Schkufza, for their insightful comments. Julia McClellan helped with early exploration of the work. ...
dblp:conf/nsdi/GuoHH21
fatcat:yf7dnug7wbcyzen7ttlxn6dhme
Exocompilation for productive programming of hardware accelerators
2022
Proceedings of the 43rd ACM SIGPLAN International Conference on Programming Language Design and Implementation
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Schedules are defined as composable rewrites within the language, and we develop a set of effect analyses which guarantee program equivalence and memory safety through these transformations. ...
are commonly still coded and optimized by hand, at great expense, in low-level C and assembly. ...
This work was partially supported by the Applications Driving Architectures (ADA) center, one of six centers of JUMP, a Semiconductor Research Corporation program co-sponsored by DARPA. ...
doi:10.1145/3519939.3523446
fatcat:ak3wlrnzvvacrktvdzjqfa5pdm
A Survey on the Optimization of Neural Network Accelerators for Micro-AI On-Device Inference
2021
IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
techniques, and the realization of the micro-AI models on resource-constrained hardware and different design considerations associated with it. ...
The efficacy of DNNs coincides with the fact that they can provide state-ofthe-art inference accuracy for these applications. ...
In this case, the transformation matrices are replaced with FFT of both the image and the kernel. ...
doi:10.1109/jetcas.2021.3129415
fatcat:nknpy4eernaeljz2hpqafe7sja
« Previous
Showing results 1 — 15 out of 1,663 results