• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# TensorFlow Lite Micro Mbed Project
2
3This folder has been autogenerated by TensorFlow, and contains source, header,
4and project files needed to build a single TensorFlow Lite Micro target using
5the Mbed command line interface.
6
7## Usage
8
9To load the dependencies this code requires, run:
10
11```
12mbed config root .
13mbed deploy
14```
15
16TensorFlow requires C++ 11, so you'll need to update your profiles to reflect
17this. Here's a short Python command that does that:
18
19```
20python -c 'import fileinput, glob;
21for filename in glob.glob("mbed-os/tools/profiles/*.json"):
22  for line in fileinput.input(filename, inplace=True):
23    print line.replace("\"-std=gnu++98\"","\"-std=c++11\", \"-fpermissive\"")'
24```
25
26With that setting updated, you should now be able to compile:
27
28```
29mbed compile -m auto -t GCC_ARM
30```
31
32If this works, it will give you a .bin file that you can flash onto the device
33you're targeting. For example, using a Discovery STM3246G board, you can deploy
34it by copying the bin to the volume mounted as a USB drive, just by dragging
35over the file.
36
37## Project Generation
38
39See
40[tensorflow/lite/micro](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/micro)
41for details on how projects like this can be generated from the main source
42tree.
43
44## License
45
46TensorFlow's code is covered by the Apache2 License included in the repository,
47and third party dependencies are covered by their respective licenses, in the
48third_party folder of this package.
49