Relay: A High-Level IR for Deep Learning
@article{Roesch2019RelayAH, title={Relay: A High-Level IR for Deep Learning}, author={Jared Roesch and Steven Lyubomirsky and Marisa Kirisame and J. Pollock and Logan Weber and Ziheng Jiang and T. Chen and T. Moreau and Zachary Tatlock}, journal={ArXiv}, year={2019}, volume={abs/1904.08368} }
Frameworks for writing, compiling, and optimizing deep learning (DL) models have recently enabled progress in areas like computer vision and natural language processing. Extending these frameworks to accommodate the rapidly diversifying landscape of DL models and hardware platforms presents challenging tradeoffs between expressiveness, composability, and portability. We present Relay, a new intermediate representation (IR) and compiler framework for DL models. The functional, statically-typed… CONTINUE READING
Figures, Tables, and Topics from this paper
5 Citations
Enhancing the interoperability between deep learning frameworks by model conversion
- Computer Science
- ESEC/SIGSOFT FSE
- 2020
- 1
- PDF
Nimble: Efficiently Compiling Dynamic Neural Networks for Model Inference
- Computer Science
- ArXiv
- 2020
- 3
- PDF
FlexTensor: An Automatic Schedule Exploration and Optimization Framework for Tensor Computation on Heterogeneous System
- Computer Science
- ASPLOS
- 2020
- 8
Ansor : Generating High-Performance Tensor Programs for Deep Learning
- Computer Science, Mathematics
- OSDI
- 2020
- 8
- PDF
References
SHOWING 1-10 OF 39 REFERENCES
ONNC: A Compilation Framework Connecting ONNX to Proprietary Deep Learning Accelerators
- Computer Science
- 2019 IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS)
- 2019
- 10
- Highly Influential
- PDF
Glow: Graph Lowering Compiler Techniques for Neural Networks
- Computer Science
- ArXiv
- 2018
- 101
- Highly Influential
- PDF
Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions
- Computer Science
- ArXiv
- 2018
- 189
- PDF
Caffe: Convolutional Architecture for Fast Feature Embedding
- Computer Science
- ACM Multimedia
- 2014
- 12,573
- PDF
AutoGraph: Imperative-style Coding with Graph-based Performance
- Computer Science, Mathematics
- ArXiv
- 2018
- 20
- PDF
Deep Learning with Dynamic Computation Graphs
- Computer Science, Mathematics
- ICLR
- 2017
- 98
- Highly Influential
- PDF
Quantizing deep convolutional networks for efficient inference: A whitepaper
- Computer Science, Mathematics
- ArXiv
- 2018
- 247
- PDF
Halide: a language and compiler for optimizing parallelism, locality, and recomputation in image processing pipelines
- Computer Science
- PLDI '13
- 2013
- 703
- Highly Influential
- PDF