posgnu / DL_Compiler

Study Group of Deep Learning Compiler

Deep Learning Compiler Study

This is a repository of the study "DL Compiler". The goal of this study is to understand the acceleration of nerual networks with DL Compiler. The topic of acceleration includes On-Device AI,DL Compiler, TVM, ONNX , Compiler. Our study is based on this paper (The Deep Learning Compiler: A Comprehensive Survey, IEEE TPDS 2021). Also we discuss other topics such as HW architecture, SW acceleration. Our materials are open to git and youtube.

Presentation with Video

TVM: An Automated End-to-End Optimizing Compiler for Deep Learning

Presenter: Constant Park ([e-mail])
Date: February, 25, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/TVM.pdf
Video: https://youtu.be/wzy1QMci_Zs

XLA: Optimizing Compiler for Machine Learning

Presenter: Tee Jung ([e-mail], https://b.mytears.org/)
Date: March, 11, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/XLA101.pdf
Video: https://youtu.be/_3ykXQH5h2o

Efficient Execution of Quantized Deep Learning Models: A Compiler Approach

Presenter: 이제민 ([e-mail])
Date: March, 25, 2021
PPT: https://www.slideshare.net/leejaymin/efficient-execution-of-quantized-deep-learning-models-a-compiler-approach
Video: https://youtu.be/JV31xwqJUKI

PlaidML: Portable Deep Learning Compiler

Presenter: Seo Sanghyeon ([e-mail])
Date: April, 22, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/PlaidML.pdf
Video: https://youtu.be/GJ_IYfVmPg4

AutoTVM and Auto Scheduler

Presenter: 류재훈 ([e-mail])
Date: April, 22, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Auto_Opt.pdf
Video: https://youtu.be/rl8pobauUn4

MLIR: A Compiler Infrastructure for the End of Moore’s Law

Presenter: Dong-hee Na ([e-mail])
Date: May, 06, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Introduction%20to%20MLIR.pdf
Video: https://youtu.be/vZy_aHERPDY

BYOC: Bring Your Own Codegen to Deep Learning Compiler

Presenter: Hyunwoo Cho
Date: May, 20, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/BYOC.pdf
Video: https://youtu.be/q3jE7nu0EgQ

Tensor Comprehension

Presenter: Jungju Oh
Date: June, 10, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Tensor%20Comprehensions.pdf
Video: https://youtu.be/8MutpjppKlw

Chameleon: Adaoptive Code Optimization for Expedited Deep Neural Network Compilation

Presenter: Taehee Jeong
Date: June, 14, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/%5BDL%20Study%5D%20Chameleon_%20Adaptive%20Code%20Optimization%20for%20Expedited%20Deep%20Neural%20Network%20Compilation.pdf
Video: https://youtu.be/vCJpEwSnEu0

Glow: Graph Lowering Compiler Techniques for Neural Networks

Presenter: Jeongho Kim
Date: July, 1, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Glow_%20Graph%20Lowering%20Compiler%20Techniques%20for%20Neural%20Networks.pdf
Video: https://youtu.be/wmIiPUDgzl4

Glow for NXP MCUs

Presenter: Dongshik Won
Date: July, 15, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Glow%20for%20NXP%20MCUs.pdf
Video: https://youtu.be/6ALFNYbnnQs	

TensorDIMM: Practical Near-Memory Processing Archiecture for Embeddings and Tensor Operations in DL

Presenter: Constant Park
Date: August, 05, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/TensorDIMM.pdf
Video: -	

ConfuciuX: Autonomous Hardware Resource Assignment for DNN Accelerators using Reinforcement Learning

Presenter: Constant Park
Date: September, 09, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/ConfuciuX.pdf
Video: https://youtu.be/XWkQQQhoBMI

Optimizing DNN Computation with Relaxed Graph Substitutions & TASO: Optimizing Deep Learning Computation with Automatic Generation of Graph Substitutions

Presenter: 류재훈 ([e-mail])
Date: September, 30, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/taso.pdf
Video: https://youtu.be/XZdnRYbM1g0

HAWQ-V3: Dyadic Neural Network Quantization

Presenter: 이제민 ([e-mail])
Date: October, 14, 2021
PPT: https://www.slideshare.net/leejaymin/hawqv3-dyadic-neural-network-quantization
Video: https://www.youtube.com/watch?v=Hxrw4cDM0Tw&list=UU03m_PqzOeNJPmZyyY2dRQw&index=2

I-BERT: Integer-only BERT Quantization

Presenter: Dongshik Won
Date: November, 11, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/I-BERT_%20Integer-only%20BERT%20Quantization.pdf
Video: https://youtu.be/--Is5DxG1wU

DNNFusion: Accelerating Deep Neural Networks Execution with Advanced Operator Fusion

Presenter: Taehee Jeong
Date: November, 15, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/DNNFusion.pdf
Video: https://youtu.be/P-LZ-RZIH0U

TENET: A Framework for Modeling Tensor Dataflow Based on Relation-centric Notation

Presenter: Hyunwoo Cho
Date: December, 09, 2021
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/DLC_Study_211209_HyunwooCho.pdf
Video: https://youtu.be/snh4BZ0v6jI

AIMET: AI Model Efficiency Toolkit

Presenter: Tee Jung ([e-mail], https://b.mytears.org/)
Date: January, 06, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/AIMET.pdf
Video: -

Newton: A DRAM-maker's Accelerator-in-Memory (AiM) Architecture for ML

Presenter: Yongwon Shin ([e-mail])
Date: February, 17, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/newton.pdf
Video: https://youtu.be/2076HWa7abY

Heterogeneous Dataflow Accelerators for Multi-DNN Workloads

Presenter:  Constant Park
Date: August, 08, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/HDA.pdf
Video: https://youtu.be/C_GyaR4ukP0

MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer

Presenter:  이제민 ([e-mail])
Date: August, 22, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/220822_MobileViTv1.pdf
Video: https://youtu.be/dVH02_O2MzQ

AsyMo: Scalable and Efficient Deep-Learning Inference on Asymmetric Mobile CPUs

Presenter:  박준형 ([e-mail])
Date: September, 19, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/AsyMo-%E1%84%80%E1%85%A9%E1%86%BC%E1%84%80%E1%85%A2%E1%84%8B%E1%85%AD%E1%86%BC.pptx.pdf
Video: https://youtu.be/MKYkq92Hdbk

Unity: Accelerating DNN Training Through Joint Optimization of Algebraic Transformations and Parallelization

Presenter:  류재훈 ([e-mail])
Date: September, 19, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/Unity.pdf
Video: https://youtu.be/YMlXaP6uHnU

Google Neural Network Models for Edge Devices: Analyzing and Mitigating Machine Learning Inference Bottlenecks

Presenter:  신용원 ([e-mail])
Date: September, 19, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/MENSA.pdf
Video: https://youtu.be/bnpdoZQB6Qs

Adaptable Butterfly Accelerator for Attention-based NNs via Hardware and Algorithm Co-design

Presenter:  Hyunwoo Cho ([e-mail])
Date: November, 11, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/DLC_Study_111422_hwcho.pdf
Video: https://youtu.be/7-aKgDxxOXU

A Learned Performance Model for Tensor Processing Units

Presenter:  박주언 ([e-mail])
Date: November, 19, 2022
PPT: -
Video: https://youtu.be/g-MJlRgRfto

The Deep Learning Compiler: A Comprehensive Survey

Presenter:  이태영 ([e-mail])
Date: December, 12, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/The%20Deep%20Learning%20Compiler.pdf
Video: https://youtu.be/O2TjOvYl8Ys	

Distilling Bit-level Sparsity Parallelism for General Purpose Deep Learning Acceleration

Presenter:  노대철 ([e-mail])
Date: November, 26, 2022
PPT: https://github.com/ConstantPark/DL_Compiler/blob/main/221226%20-%20Distilling%20Bit-level%20Sparsity%20Parallelism%20for%20General%20Purpose%20Deep%20Learning%20Acceleration.pdf
Video: https://youtu.be/7-aKgDxxOXU	

Alpa: Automating Inter- and Intra-Operator Parallelism for Distributed Deep Learning

Presenter:  윤유경 ([e-mail])
Date: Januaray, 09, 2023
PPT: -
Video: https://youtu.be/LGYYRRKxCjE