The nest-compiler from etri

nest-compiler's Introduction

Partners

Contributions to NEST-C are welcomed and encouraged! We are proud to collaborate with the following companies on our open-source project:

We appreciate their support and contributions to our project. Click on the logos to learn more about each partner and their involvement with our project.

Contacts

PI

[email protected]

Developers

Overview

The NEST Compiler (NEST-C) is an open source project led by ETRI, which is based on [GLOW project] (https://github.com/pytorch/glow).

Glow is a machine learning compiler and execution engine for hardware accelerators. It is designed to be used as a backend for high-level machine learning frameworks. The compiler is designed to allow state of the art compiler optimizations and code generation of neural network graphs.

The objective of NEST-C is to generate optimized code for various kinds of Neural-network Processing Uints (NPUs). Therefore, NEST-C provides automatic tuning functionalities and tools for each optimization step.

Optimization modules and tools of NEST-Compiler

Input/output examples of NEST-C modules

Features

NPU HW backend supporting EVTA series
Automatic backend parameter tuning (under development)
Profile-based dyanamic quantization for supporting various NPU backends
C/C++ code generation for embedded HWs supporting gcc/llvm compilers
Profile-based graph partitioning

Documentation

Research Papers

PartitionTuner: An operator scheduler for deep-learning compilers supporting multiple heterogeneous processing units
- ETRI Journal Vol. 45 No. 2 (JCR21 IF: 1.622) ISSN: 1225-6463, doi: https://doi.org/10.4218/etrij.2021-0446, Apr 2023.
Quantune: Post-training quantization of convolutional neural networks using extreme gradient boosting for fast deployment
- Future Generation Computer Systems Vol. 132, 2022, pp. 124-135, July 01, 2022 (JCR21 IF: 7.307, Top 9.09%), Jul 2022
범용 AI 컴파일러의 비공개 NPU 코드생성을 위한 공통 인터페이스 설계 및 검증
- 전자공학회논문지, v.60 no.10, pp.29-31, 2023.10.25, 우수등재지

Demo

License

NEST-C is licensed under the Apache 2.0 License.

nest-compiler's People

Contributors

Stargazers

Watchers

nest-compiler's Issues

I have question at testing and running | installation

on installation document ...

Configure and Build
To build the compiler, create a build directory and run cmake on the source directory. It's a good idea to build two configurations (Release and Debug) because some programs take a really long time to run in Debug mode. It's also a good idea to build the project outside of the source directory.

mkdir build
cd build
export TVM_HOME=/path/to/nest_compiler/tvm
export PYTHONPATH=/path/to/nest_compiler/tvm/python
export TVM_LIBRARY_PATH=/path/to/build/tvm
cmake -G Ninja /path/to/nest_compiler -DCMAKE_BUILD_TYPE=Release -DNESTC_WITH_EVTA=ON -DNESTC_EVTA_BUNDLE_TEST=ON -DGLOW_WITH_BUNDLES=ON -DNESTC_USE_PRECOMPILED_EVTA_LIBRARY=ON
ninja check_nestc

Where should I put that build folder?
Also, where exactly should I route the TVM?
Should I grab it as HOME/USER?
Also, when I run cmake, I get some errors, but there is no requirement.txt of the project?

on testing and running document...

Now run the examples. Note that the databases should be in the current working directory.

./bin/mnist
./bin/cifar10
./bin/fr2en
./bin/ptb
./bin/char-rnn

where is that ./bin directory...?
that download_datasets_and_models.py --all-datasets dont have bin...

Recommend Projects

etri / nest-compiler Goto Github PK