5 Hits in 3.6 sec

ProTuner: Tuning Programs with Monte Carlo Tree Search [article]

Ameer Haj-Ali, Hasan Genc, Qijing Huang, William Moses, John Wawrzynek, Krste Asanović, Ion Stoica
2020 arXiv   pre-print
We explore applying the Monte Carlo Tree Search (MCTS) algorithm in a notoriously difficult task: tuning programs for high-performance deep learning and image processing.  ...  We further explore modifications to the standard MCTS algorithm as well as combining real execution time measurements with the cost model.  ...  Monte Carlo Tree Search Monte Carlo Tree Search (MCTS) is a method that can solve MDPs. It combines tree search with random sampling for finding the optimal decisions in the MDP.  ... 
arXiv:2005.13685v1 fatcat:3yrh5sxgbrfojgcjsjfqcylnvi

Customized Monte Carlo Tree Search for LLVM/Polly's Composable Loop Optimization Transformations [article]

Jaehoon Koo, Prasanna Balaprakash, Michael Kruse, Xingfu Wu, Paul Hovland, Mary Hall
2021 arXiv   pre-print
We have developed a search algorithm based on Monte Carlo tree search (MCTS) to find the best combination of loop transformations.  ...  We compare our approach with random, greedy, and breadth-first search methods on PolyBench kernels and ECP proxy applications.  ...  Monte Carlo tree search Monte Carlo tree search (MCTS) is a heuristic search method that seeks to solve a class of computationally intractable sequential decision-making problems, typically represented  ... 
arXiv:2105.04555v1 fatcat:jiduf3iel5fbfgmx5ni3ugapnq

Ansor : Generating High-Performance Tensor Programs for Deep Learning [article]

Lianmin Zheng, Chengfan Jia, Minmin Sun, Zhao Wu, Cody Hao Yu, Ameer Haj-Ali, Yida Wang, Jun Yang, Danyang Zhuo, Koushik Sen, Joseph E. Gonzalez, Ion Stoica
2020 arXiv   pre-print
Ansor then fine-tunes the sampled programs with evolutionary search and a learned cost model to identify the best programs.  ...  Compared with existing search strategies, Ansor explores many more optimization combinations by sampling programs from a hierarchical representation of the search space.  ...  A concurrent work ProTuner [19] uses Monte Carlo tree search to solve the inaccurate estimation problem in Halide auto-scheduler.  ... 
arXiv:2006.06762v4 fatcat:as6rrj2bvjcwtmkjremrrfkqhq

Autotuning PolyBench Benchmarks with LLVM Clang/Polly Loop Optimization Pragmas Using Bayesian Optimization (extended version) [article]

Xingfu Wu, Michael Kruse, Prasanna Balaprakash, Hal Finkel, Paul Hovland, Valerie Taylor, Mary Hall
two large datasets in 200 code evaluations for effectively searching the parameter spaces with up to 170,368 different configurations.  ...  We also extend the ytopt autotuning framework to tune a deep learning application.  ...  ProTuner 49 further improved Halide schedule autotuning by replacing Beam search with Monte Carlo tree search.  ... 
doi:10.48550/arxiv.2104.13242 fatcat:x4rdwdmi4vcnfkrtqcg4v2pqaq

Biennial Report 2012-2013 [article]

CSIC-Instituto De Química Avanzada De Cataluña (IQAC), Digital.CSIC, Digital.CSIC
The spectrometer has a VNMRS console with Direct Digital receiver with Performa-I Z gradients and OneNMR probe, which can be tuned automatically to any nucleus between 15 N and 31 P with ProTune System  ...  Colloid & Interface Sci., 17, 255-260, 2012 ························ Brownian dynamics and dynamic Monte Carlo simulations of isotropic and liquid crystal phases of anisotropic colloidal particles: A comparative  ...  Immunization, cell fusion, screening and cloning: After their fusion with myeloma cells screening of the best hybridomas will be performed by ELISA using previously supplied antigen.  ... 
doi:10.20350/digitalcsic/10756 fatcat:7sf7lhhlynfo5p5mnpd5rmgqg4