edoffagne/utensor_cgen

C++ code generator for uTensor

Installation (Python 2 & 3)

• installation with setup.py

python setup.py install

• installation with pip

pip install utensor_cgen

Develop Environment

We use pipenv to setup the develop environment.

You can go to this repo for detail information about pipenv.

Setup with pipenv

1. # pipenv install -d
2. # pipenv shell
   • this will spawn a subshell and activate the virtual environment for you
   • You should be able to use the cli now

Example

Please refer to tests/deep_mlp for detailed example

User Guild

Following steps are a general guild for user how to porting a TensorFlow protobuf file into a uTensor implementation:

1. Freeze and quantize your graph
   • Freezing
   • Quantization
     • An alternative is to use the quantize_graph.py script
     • it should output one qunatized pb file, say quantized_graph.pb
2. install utensor_cgent
   • run python3 setupt.py install
3. run utensor-cli quantized_graph.pb, where quantized_graph.pb is the output pb file you get from step 1
   • run utensor-cli -h for help

TODOs

1. (done) Freezed graph protobuff parser
2. (done)Tensor snippets for uTensor
3. (done) Add template engine for richer flexibility
   • jinja2
4. (done?) core code generator implementation
   • We need some refatoring, PRs are welcomed!
5. type alias in C/C++
   • ex: use uint8_t or unsigned char?
   • a lot more about this....
6. (done) MANIFAST.in for the setup.py
7. Relation among snippets/containers
   • shared template variables? (headers, shared placeholders...etc)
8. Better configuration schema
   • json
   • yaml
   • or ?