https://github.com/xiezhq-hermann/tinyinfer
https://github.com/xiezhq-hermann/tinyinfer
Last synced: 7 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/xiezhq-hermann/tinyinfer
- Owner: xiezhq-hermann
- Created: 2019-05-27T03:13:43.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-05-31T07:08:10.000Z (over 6 years ago)
- Last Synced: 2025-01-28T20:12:24.661Z (9 months ago)
- Language: Jupyter Notebook
- Size: 1.33 MB
- Stars: 0
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# TinyInfer
## Dependency
#### Libraries
- [google protobuf](https://github.com/protocolbuffers/protobuf)
- [nlohmann json](https://github.com/nlohmann/json)
- eigen3 (auto-download on building)
Notes:
- Please make sure your protobuf library supports proto3
- It is recommended to compile and install protobuf and nlohmann-json from source code
#### Local Test Environment
- Ubuntu 16.04
- Protobuf 3.7.1
## Usage
### Compilation
```bash
mkdir build
cd build
cmake ..
make -j3
```
### using for MNIST inference
#### Data
- download data from the [release page](https://github.com/jianxiongcai/cs133_tinyinfer_data/releases) of the data repo.
- untar the file in project root directory. This will have everything needed for running demo in ```data``` folder.
- Alternatively you may refer to the jupyter notebooks under [train/jupyter](train/jupyter) to generate new input data.
#### Generate input tensor and weights
Refer to jupyter nodebooks under [demo/jupyter/demo.ipynb](demo/jupyter/demo.ipynb)
#### Inference
```bash
./predict {PATH_TO_PB_MODEL} {DIR_TO_TENSOR_WEIGHTS} {PATH_TO_INPUT}
# for exmaple
./predict ../data/model.pb ../data/tensor_weights ../data/demo_input/input.tensor
```
Alternatively, we provide a jupyter notebook for reference. (See jupyter Notebook [demo.ipynb](demo/jupyter/demo.ipynb))
### static library
```cmake
add_subdirectory(src)
```
This will invoke the cmake file in src and produce a static library ```TinyinferLib```, which can be included for inferring other network.
example usage:
```cmake
add_executable(predict test/predict.cpp)
target_link_libraries(predict TinyinferLib)
```
### Acknowledgement
This project is the course project for CS133 advanced C++ at ShanghaiTech University.
### Presentation Archive
see [CS133-Presentation.pdf](CS133-Presentation.pdf)
### Author
- Jianxiong Cai
- Zhiqiang Xie
- Jiadi Cui