https://github.com/modeltc/pyvlova
Yet another Polyhedra Compiler for DeepLearning
https://github.com/modeltc/pyvlova
Last synced: 7 months ago
JSON representation
Yet another Polyhedra Compiler for DeepLearning
- Host: GitHub
- URL: https://github.com/modeltc/pyvlova
- Owner: ModelTC
- License: apache-2.0
- Created: 2020-08-14T05:50:24.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2023-04-14T17:28:07.000Z (almost 3 years ago)
- Last Synced: 2025-04-04T21:39:41.672Z (11 months ago)
- Language: Python
- Size: 177 KB
- Stars: 19
- Watchers: 4
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Pyvlova
A Simple Polyhedral Compiler for NN
# Benchmark
- [ResNet-18](doc/benchmark/resnet18.md)
- [MobileNetV2](doc/benchmark/mobilenetv2.md)
# Requirements
- `tvm`
- `sympy`, `numpy`
# Build extensions (Build ISL)
```bash
git submodule update --init --recursive --progress 3rdparty/isl
python setup.py build_ext
```
## ISL
- `isl` ([ChieloNewctle/isl](https://github.com/ChieloNewctle/isl))
### Requirements
- `llvm` and `clang`
- `libgmp-dev`
- `libclang-*-dev`, such as `libclang-10-dev`, **IMPORTANT for building the python interface**
- `automake`, `autoconf`, `libtool`
- `pkg-config`
# Run tests
To test CUDA kernels generated by pyvlova:
```bash
python -m unittest test.test_op
```