mlsanityData to 7 Oct 2026

Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions

arXiv 1802.04730 ↗First posted 13 Feb 2018cs.PL cs.LGcited by 84

Not on the map. Why?

Abstract (arXiv)

Deep learning models with convolutional and recurrent networks are now ubiquitous and analyze massive amounts of audio, image, video, text and graph data, with applications in automatic translation, speech-to-text, scene understanding, ranking user preferences, ad placement, etc. Competing frameworks for building these networks such as TensorFlow, Chainer, CNTK, Torch/PyTorch, Caffe1/2, MXNet and Theano, explore different tradeoffs between usability and expressiveness, research or production orientation and supported hardware. They operate on a DAG of computational operators, wrapping high-performance libraries such as CUDNN (for NVIDIA GPUs) or NNPACK (for various CPUs), and automate memory allocation, synchronization, distribution. Custom operators are needed where the computation does not fit existing high-performance library calls, usually at a high engineering cost. This is frequently required when new operators are invented by researchers: such operators suffer a severe performance penalty, which limits the pace of innovation. Furthermore, even if there is an existing runtime call these frameworks can use, it often doesn't offer optimal performance for a user's particular network architecture and dataset, missing optimizations between operators as well as optimizations that can be done knowing the size and shape of data. Our contributions include (1) a language close to the mathematics of deep learning called Tensor Comprehensions, (2) a polyhedral Just-In-Time compiler to convert a mathematical description of a deep learning DAG into a CUDA kernel with delegated memory management and synchronization, also providing optimizations such as operator fusion and specialization for specific sizes, (3) a compilation cache populated by an autotuner. [Abstract cutoff]

Key facts. Cited by 84 papers in the map (data through 7 Oct 2026). First posted on arXiv 13 Feb 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 19; 2022: 13; 2023: 16; 2024: 16; 2025: 9; 2026: 11.

84 papers.

Sort:

By area:

  1. 2021Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Bettercited by 153
  2. 2023Full Stack Optimization of Transformer Inference: a Surveycited by 79
  3. 2021DNNFusion: Accelerating Deep Neural Networks Execution with Advanced Operator Fusioncited by 56
  4. 2022TensorIR: An Abstraction for Automatic Tensorized Program Optimizationcited by 38
  5. 2024ThunderKittens: Simple, Fast, and Adorable AI Kernelscited by 34
  6. 2021A Deep Learning Based Cost Model for Automatic Code Optimizationcited by 33
  7. 2022Tensor Program Optimization with Probabilistic Programscited by 30
  8. 2022Hidet: Task-Mapping Programming Paradigm for Deep Learning Tensor Programscited by 28
  9. 2021Mind Mappings: Enabling Efficient Algorithm-Accelerator Mapping Space Searchcited by 24
  10. 2021CoSA: Scheduling by Constrained Optimization for Spatial Acceleratorscited by 22
  11. 2021The CoRa Tensor Compiler: Compilation for Ragged Tensors with Minimal Paddingcited by 17
  12. 2022Tensor and Matrix Low-Rank Value-Function Approximation in Reinforcement Learningcited by 16
38 more
  1. 2023TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphscited by 16
  2. 2024Allo: A Programming Model for Composable Accelerator Designcited by 16
  3. 2021MAGMA: An Optimization Framework for Mapping Multiple DNNs on Multiple Accelerator Corescited by 15
  4. 2022TLP: A Deep Learning-based Cost Model for Tensor Program Tuningcited by 14
  5. 2021UNIT: Unifying Tensorized Instruction Compilationcited by 10
  6. 2021HASCO: Towards Agile HArdware and Software CO-design for Tensor Computationcited by 10
  7. 2023GraphRARE: Reinforcement Learning Enhanced Graph Neural Network with Relative Entropycited by 9
  8. 2021Neural Architecture Search as Program Transformation Explorationcited by 8
  9. 2021Tensor Processing Primitives: A Programming Abstraction for Efficiency and Portability in Deep Learning & HPC Workloadscited by 8
  10. 2023Operator Fusion in XLA: Analysis and Evaluationcited by 8
  11. 2021Horizontally Fused Training Array: An Effective Hardware Utilization Squeezer for Training Novel Deep Learning Modelscited by 7
  12. 2022Decompiling x86 Deep Neural Network Executablescited by 7
  13. 2023oneDNN Graph Compiler: A Hybrid Approach for High-Performance Deep Learning Compilationcited by 7
  14. 2024LOOPer: A Learned Automatic Code Optimizer For Polyhedral Compilerscited by 7
  15. 2021Functional Collection Programming with Semi-Ring Dictionariescited by 6
  16. 2021A Reinforcement Learning Environment for Polyhedral Optimizationscited by 5
  17. 2021Bring Your Own Codegen to Deep Learning Compilercited by 5
  18. 2024SPLAT: A framework for optimised GPU code-generation for SParse reguLar ATtentioncited by 5
  19. 2024Scaling Deep Learning Computation over the Inter-Core Connected Intelligence Processor with T10cited by 5
  20. 2025From Tiny Machine Learning to Tiny Deep Learning: A Surveycited by 5
  21. 2021A Highly Configurable Hardware/Software Stack for DNN Inference Accelerationcited by 4
  22. 2022Neural Architecture Search using Property Guided Synthesiscited by 4
  23. 2023Harnessing Deep Learning and HPC Kernels via High-Level Loop and Tensor Abstractions on CPU Architecturescited by 4
  24. 2024Unraveling the Key of Machine Learning-based Android Malware Detectioncited by 4
  25. 2024SoD2: Statically Optimizing Dynamic Deep Neural Networkcited by 4
  26. 2024Dynamic Co-Optimization Compiler: Leveraging Multi-Agent Reinforcement Learning for Enhanced DNN Accelerator Performancecited by 4
  27. 2025Agentic Auto-Scheduling: An Experimental Study of LLM-Guided Loop Optimizationcited by 4
  28. 2021A Data-Centric Optimization Framework for Machine Learningcited by 3
  29. 2022Benchmarking the Linear Algebra Awareness of TensorFlow and PyTorchcited by 3
  30. 2024CIM-MLC: A Multi-level Compilation Stack for Computing-In-Memory Acceleratorscited by 3
  31. 2024A Reinforcement Learning Environment for Automatic Code Optimization in the MLIR Compilercited by 3
  32. 2025Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computationcited by 3
  33. 2025REASONING COMPILER: LLM-Guided Optimizations for Efficient Model Servingcited by 3
  34. 2021Collage: Seamless Integration of Deep Learning Backends with Automatic Placementcited by 2
  35. 2022Demystifying Map Space Exploration for NPUscited by 2
  36. 2022Legal-Tech Open Diaries: Lesson learned on how to develop and deploy light-weight models in the era of humongous Language Modelscited by 2
  37. 2023Saturn: An Optimized Data System for Large Model Deep Learning Workloadscited by 2
  38. 2023Serving Deep Learning Model in Relational Databasescited by 2

50 most cited of 84 All 84 (JSON)