first commit

.gitignore

@@ -0,0 +1,4 @@
+try.py
+venv
+.idea
+tinymlgen/__pycache__

README.md

@@ -0,0 +1,32 @@
+# TinyML gen
+
+This is a simple package to export a model trained in Tensorflow Lite
+to a plain C array, ready to be used for inference on microcontrollers.
+
+### Install
+
+```shell script
+pip install tinymlgen
+```
+
+### Use
+
+```python
+from tinymlgen import port
+
+if __name__ == '__main__':
+    tf_model = create_tf_model()
+    c_code = port(tf_model)
+```
+
+### Configuration
+
+You can pass a few parameters to the `port` function:
+
+ - `optimize (=True)`: apply optimizers to the exported model.
+    Can either be a list of optimizers or a boolean, in which case
+    `OPTIMIZE_FOR_SIZE` is applied
+ - `variable_name (='model_data')`: give the exported array a custom name
+ - `pretty_print (=False)`: print the array in a nicely formatted arrangement
+
+

setup.py

@@ -0,0 +1,27 @@
+from distutils.core import setup
+setup(
+  name = 'tinymlgen',
+  packages = ['tinymlgen'],
+  version = '0.1',
+  license='MIT',
+  description = 'Generate C code for microcontrollers from Tensorflow models',
+  author = 'Simone Salerno',
+  author_email = '[email protected]',
+  url = 'https://github.com/eloquentarduino/tinymlgen',
+  download_url = 'https://github.com/eloquentarduino/tinymlgen/archive/v_01.tar.gz',
+  keywords = ['ML', 'microcontrollers', 'tensorflow', 'machine learning'],
+  install_requires=[
+    'tensorflow',
+    'hexdump'
+  ],
+  classifiers=[
+    'Development Status :: 3 - Alpha',
+    'Intended Audience :: Developers',
+    'Topic :: Software Development :: Code Generators',
+    'License :: OSI Approved :: MIT License',
+    'Programming Language :: Python :: 3',
+    'Programming Language :: Python :: 3.4',
+    'Programming Language :: Python :: 3.5',
+    'Programming Language :: Python :: 3.6',
+  ],
+)

tinymlgen/tinymlgen.py

@@ -0,0 +1,35 @@
+import re
+import hexdump
+import tensorflow as tf
+
+
+def port(model, optimize=True, variable_name='model_data', pretty_print=False):
+    converter = tf.lite.TFLiteConverter.from_keras_model(model)
+    if optimize:
+        if isinstance(optimize, bool):
+            optimizers = [tf.lite.Optimize.OPTIMIZE_FOR_SIZE]
+        else:
+            optimizers = optimize
+        converter.optimizations = optimizers
+    tflite_model = converter.convert()
+    bytes = hexdump.dump(tflite_model).split(' ')
+    c_array = ', '.join(['0x%02x' % int(byte, 16) for byte in bytes])
+    c = 'const unsigned char %s[] DATA_ALIGN_ATTRIBUTE = {%s};' % (variable_name, c_array)
+    if pretty_print:
+        c = c.replace('{', '{\n\t').replace('}', '\n}')
+        c = re.sub(r'(0x..?, ){12}', lambda x: '%s\n\t' % x.group(0), c)
+    c += '\nconst int %s_len = %d;' % (variable_name, len(bytes))
+    preamble = '''
+#ifdef __has_attribute
+#define HAVE_ATTRIBUTE(x) __has_attribute(x)
+#else
+#define HAVE_ATTRIBUTE(x) 0
+#endif
+#if HAVE_ATTRIBUTE(aligned) || (defined(__GNUC__) && !defined(__clang__))
+#define DATA_ALIGN_ATTRIBUTE __attribute__((aligned(4)))
+#else
+#define DATA_ALIGN_ATTRIBUTE
+#endif
+
+'''
+    return preamble + c