Apache TVM and Deep Learning Compilation Conference

Thu-Fri, December 3rd-4th 2020, Virtual.

Apache (incubating) TVM is an open-source deep learning compiler stack for CPUs, GPUs, and specialized accelerators. It aims to close the gap between the productivity-focused deep learning frameworks, and the performance- or efficiency-oriented hardware backends.

The 2019 TVM Conference covered the state of the art of deep learning compilation optimization. TVM contributors, users, UW SAMPL sponsors, collaborators and researchers and practitioners from the broader community met in Seattle to discuss recent advances in frameworks, compilers, systems and architecture support, security, training and hardware acceleration.

2019 Program

Presentation slides and video recordings are now available.

9:00 Keynote and Community Update – Luis Ceze, Jeff Gehlhaar, Yida Wang, Zach Tatlock, Jason Knight, Tianqi Chen [Video] [Slides]  
10:00 TVM @ AWS – Yida Wang and Zhi Chen, AWS [Video] [Slides]  
10:40 TVM @ FB – Andrew Tulloch and Bram Wasti, Facebook [Video] [Slides]  
11:10 break  
11:30 AI Compilers at Alibaba – Xiaoyong Liu, Alibaba [Video] [Slides]  
12:00 Dynamic Execution and Virtual Machine, Jared Roesch and Haichen Shen, UW and AWS [Video] [Slides]  
12:20 Lunch started(boxed lunches will be provided), contributors meetup  
13:10 Lunch Lightning talk session  
  Techniques for Fast End-to-End Binarized Networks – Josh Fromm, OctoML and UW [Video] [Slides]  
  TensorCore and Tensorization – Siyuan Feng, SJTU and UW [Video] [Slides]  
  Automatic TensorCore Scheduling – Xiaoyong Liu, Alibaba [Video] [Slides]  
  Dynamic Model - Graph Dispatching – Yao Wang, AWS [Video] [Slides]  
  Efficient quantized inference on CUDA with TVM – Wuwei Lin, CMU [Video] [Slides]  
  uTVM: TVM on bare-metal devices – Logan Weber, UW [Video] [Slides]  
13:40 Building FPGA-Targeted Accelerators with HeteroCL – Zhiru Zhang, Cornell [Video] [Slides]  
14:10 TVM @ Microsoft – Jon Soifer and Minjia Zhang [Video] [Slides]  
14:30 TVM @ ARM – Ramana Radhakrishnan [Video] [Slides]  
14:50 TVM @ Xilinx – Elliott Delaye [Video] [Slides]  
15:10 break  
15:30 TVM @ OctoML – Jason Knight [Video] [Slides]  
15:50 TVM @ Qualcomm – Krzysztof Parzyszek [Video] [Slides]  
16:10 TASO: Optimizing Deep Learning Computation with Automated Generation of Graph Substitutions – Zhihao Jia, Stanford [Video] [Slides]  
16:30 Towards cross-domain co-optimization – Nilesh Jain, Intel Labs  
16:50 break  
17:00 Lightning talks session  
  Graph Convolutional Cost Models for TVM – Eddie Yan, UW [Video] [Slides]  
  Janus: Fast and Flexible Deep Learning via Symbolic Graph Execution of Imperative Deep Learning Programs – Eunji Jeong, Seoul National University [Video] [Slides]  
  Towards an operational model of schedules in Tensor Expression – Yuan Lin and Yongfeng Gu, NVIDIA [Video] [Slides]  
  Fireiron – A Scheduling Language for GPUs – Vinod Grover, NVIDIA [Video] [Slides]  
  TVM for ads ranking stack, opportunities and challenges – Hao Lu and Ansha Yu, Facebook [Video] [Slides]  
  TVM for edge computing platforms – Morita Kazutaka, NTT [Video] [Slides]  
  Improving AutoTVM Efficiency by Schedule Sharing – Cody Yu, AWS [Video] [Slides]  
  Supporting TVM on RISC-V Architectures with SIMD Computations – Jenq-Kuen Lee, NTHU [Video] [Slides]  
  Optimizing sparse/graph kernels via TVM – Yuwei Hu, Cornell and AWS [Video] [Slides]  
  Integrating model pre-processing functionality into TVM – Abelardo Lopez-Lagunas, Latent AI [Video] [Slides]  
  Deep Learning Program Analysis with Relay – Gus Smith, UW [Video] [Slides]  
  Tapasco: Task-Parallel System Composer for FPGAs – Florian Stock, TU Darmstadt [Video] [Slides]  
  Compiling Classical ML Pipelines into Tensor Computations for One-size-fits-all Prediction Serving – Matteo Interlandi, Microsoft [Video] [Slides]  
  DRAM access reduction by node fusion with TVM – Chia-Wei Chang, NTHU [Video] [Slides]  
18:10 to 20:00 Social (drinks, food)  

Follow us on Twitter

Past Conferences