vee1e/tflite-micro
1
0
Fork 0
mirror of https://github.com/vee1e/tflite-micro.git synced 2026-09-01 17:57:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Allow using Python scripts without bazel. (#639)

Browse source 
* Allow using Python scripts without bazel.

Manually verified and documented the steps that I can follow to run
Python code without Bazel.

BUG=http://b/204109200
NO_CHECK_TFLITE_FILES=manually ran the sync script.

* Address review comments.
This commit is contained in:
Advait Jain 2021-10-29 14:40:24 -07:00 • committed by GitHub
parent 9452ef8381
commit e958bb1161
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
15 changed files with 78 additions and 16 deletions
Download patch file Download diff file Expand all files Collapse all files

60
docs/python.md Normal file
View file

@ -0,0 +1,60 @@
<!--ts-->
* [Using Bazel](#using-bazel)
* [Manual Setup Illustration](#manual-setup-illustration)
<!-- Added by: advaitjain, at: Fri Oct 29 11:57:14 AM PDT 2021 -->
<!--te-->
Writing and using Python scripts from the TFLM repository is currently in the
prototyping stage. As such, the instructions below are somewhat sparse and
subject to change.
* [TensorFlow Python style guide](https://www.tensorflow.org/community/contribute/code_style#python_style)
# Using Bazel
We use Bazel as our default build system for Python and the continuous
integration infrastrucutre only runs the Python unit tests via Bazel.
When using Bazel + Python, all the environment setup is handled as part of the
build.
Some example commands:
```sh
bazel test tensorflow/lite/tools:flatbuffer_utils_test
bazel build tensorflow/lite/tools:visualize
bazel-bin/tensorflow/lite/tools/visualize tensorflow/lite/micro/models/person_detect.tflite tensorflow/lite/micro/models/person_detect.tflite.html
```
# Manual Setup Illustration
For advanced users that would like to use the Python code in the TFLM repository
independent of bazel, here is one approach.
Please note that this setup is unsupported and will need users to debug various
issues on their own. It is described here for illustrative purposes only.
```sh
# The cloned tflite-micro folder needs to be renamed to tflite_micro
mv tflite-micro tflite_micro
python -m venv tflite_micro/venv
echo "export PYTHONPATH=\${PYTHONPATH}:${PWD}" >> tflite_micro/venv/bin/activate
cd tflite_micro
source venv/bin/activate
pip install --upgrade pip
pip install -r third_party/requirements.txt
# (Optional)
pip install ipython
```
Run some tests and binaries:
```sh
python tensorflow/lite/tools/flatbuffer_utils_test.py
python tensorflow/lite/tools/visualize.py tensorflow/lite/micro/models/person_detect.tflite tensorflow/lite/micro/models/person_detect.tflite.html
```