Not presently working on windows, hopefully that changes. Actually presently I can't even get this working on Ubuntu. Try the ipython https://colab.research.google.com/drive/1OgCcKhklL3EH_SdWHdtlb5dbtYvjGQnn?usp=sharing or gitpod options https://gitpod.io/#github.com/hpssjellis/Gitpod-auto-tensorflowJS-to-arduino
- Go to the location you downloaded the model .json and model.bin files from https://hpssjellis.github.io/tinyMLjs/public/index.html
- Instal venv
sudo apt install python3.12-venvso you can work in an environment and not have other things mess it up - Install a virtual environment
python3 -m venv myenv10 - Activate that environment (note it has a folder you can stay out of )
source myenv10/bin/activateor on windowsmyenv10\scripts\activate pip install tensorflowpip install tensorflowjspip install tensorRTtflite_convert --helptensorflowjs_converter --help10.tensorflowjs_converter --input_format=tfjs_layers_model --output_format=keras_saved_model ./model.json ./Convert tfjs file to kerastflite_convert --keras_model_file ./ --output_file ./model.tfliteConvert Keras file to tflite filexxd -i model.tflite model.hConvert tflite file to a c-header file (This needs xxd installed, several ways to do this also can do it from a web page)
Non of the above work for me that is why I have the other options
- use python notebooks https://colab.research.google.com/drive/1OgCcKhklL3EH_SdWHdtlb5dbtYvjGQnn?usp=sharing (run both sketches then upload your files and run the last sketch again)
- This repo https://github.com/hpssjellis/Gitpod-auto-tensorflowJS-to-arduino and run the gitpod which has a file that does the conversions. Basically it installs the above and then you can run the commands or a bash file I have ready to do the conversions for you. The autoloading gitpod is here https://gitpod.io/#github.com/hpssjellis/Gitpod-auto-tensorflowJS-to-arduino .
.
.
This github repository is at https://github.com/hpssjellis/tensorflowjs-to-arduino-for-tinymljs
if you want to load it as a Gitpod click https://gitpod.io/#github.com/hpssjellis/tensorflowjs-to-arduino-for-tinymljs
Click the above link to load the gitpod (docker in the browser) which installs all needed files
You can use the test-with folder to drag a model.json file with it's shard .bin file model.weights.bin to the main folder
Look at the code in the a01-convert-tfjs-arduino.sh and then run it
Then run ./a01-convert-tfjs-arduino.sh
Gotchas When making your own files the model.json file is made with a link to the model.weights.bin file, if you change the name of the binary file the model.json fle will not link to it properly
I assume python is installed probably best to have Python3 installed.
pip install tensorflowjs
python -m site --user-base
to the above reply add
\bin\tensorflowjs_converter -h
Then run the commands for your files which are in the a01-convert-tfjs-arduino.sh bash file
tensorflowjs_converter --input_format=tfjs_layers_model --output_format=keras_saved_model ./model.json ./
tflite_convert --keras_model_file ./ --output_file ./model.tflite
xxd -i model.tflite model.h
Then you can load your model.tflite file onto the https://netron.app/ website to visualize it and then add the model.h file into your arduino machine learning code as it;s own include file.
See the Arduino ready library at https://github.com/hpssjellis/RocksettaTinyML download the zip file and install it into the arduino ide using the normal zip file libary upload method. sketch --> include library --> add .zip file
Note: If the above commands don't work you can always try the python code below.
- convert from tensorflowJS to Keras
import tensorflowjs as tfjs
# Define the paths
input_format = "tfjs_layers_model"
output_format = "keras_saved_model"
input_model_json = "./model.json"
output_dir = "./"
# Convert the model
tfjs.converters.save_keras_model(input_model_json, output_dir, input_format, output_format)
-
Then use tensorflow lite converter to convert the Keras file into tensorflow Lite (TFLITE)
import tensorflow as tf
# Define the paths
keras_model_file = "./model" # Make sure the model file has the .h5 extension
output_file = "./model.tflite"
# Convert the model to TensorFlow Lite
converter = tf.lite.TFLiteConverter.from_keras_model_file(keras_model_file)
tflite_model = converter.convert()
# Save the TensorFlow Lite model to a file
with open(output_file, 'wb') as f:
f.write(tflite_model)
If you have install ability then install the xxd application
sudo apt-get install xxd
and run the command
xxd -i model.tflite model.h
If you don't have admin access you can try using the online xxd -1 utility here https://hpssjellis.github.io/tinyMLjs/public/convert/xxd-i.html
Upload your tFLITE file and get the web to convert it into a c-header model.h file ready to run on a micro-controler with an appropriate sketch.
python.exe -m pip install --upgrade pip
pip3 install --upgrade pip
python -m venv myenv2
myenv2\scripts\activate
pip3 install tensorflowjs
pip3 install tensorflow==2.15.0
pip3 install tensorflow-hub
pip3 install netron "dask[delayed]"
$env:TF_ENABLE_ONEDNN_OPTS=0
tflite_convert --help tensorflowjs_converter --help
xxd --help
https://sourceforge.net/projects/xxd-for-windows/
in power shell try
Format-Hex '.\your-file-name'
this set works
pip list
--------------------------------- ---------
absl-py 2.1.0
argon2-cffi 21.1.0
astroid 2.7.3
astunparse 1.6.3
attrs 21.2.0
autopep8 1.5.7
backcall 0.2.0
backports.entry-points-selectable 1.1.0
bandit 1.7.0
bleach 4.1.0
cached-property 1.5.2
cachetools 5.3.3
certifi 2021.5.30
cffi 1.14.6
charset-normalizer 2.0.4
chex 0.1.7
click 8.1.7
cloudpickle 3.0.0
colorama 0.4.4
cryptography 3.4.8
dask 2023.5.0
debugpy 1.4.3
decorator 5.1.0
defusedxml 0.7.1
distlib 0.3.2
dm-tree 0.1.8
docutils 0.17.1
entrypoints 0.3
etils 1.3.0
filelock 3.0.12
flake8 3.9.2
flatbuffers 24.3.25
flax 0.7.2
fsspec 2024.5.0
gast 0.4.0
gitdb 4.0.7
GitPython 3.1.18
google-auth 2.29.0
google-auth-oauthlib 1.0.0
google-pasta 0.2.0
grpcio 1.63.0
h5py 3.11.0
idna 3.2
importlib_metadata 7.1.0
importlib_resources 6.4.0
ipykernel 6.4.1
ipython 7.27.0
ipython-genutils 0.2.0
isort 5.9.3
jax 0.4.13
jaxlib 0.4.13
jedi 0.18.0
jeepney 0.7.1
Jinja2 3.0.1
jsonschema 3.2.0
jupyter-client 7.0.2
jupyter-core 4.7.1
jupyterlab-pygments 0.1.2
keras 2.13.1
keyring 23.2.1
lazy-object-proxy 1.6.0
libclang 18.1.1
locket 1.0.0
Markdown 3.6
markdown-it-py 3.0.0
MarkupSafe 2.1.5
matplotlib-inline 0.1.3
mccabe 0.6.1
mdurl 0.1.2
mistune 0.8.4
ml-dtypes 0.2.0
msgpack 1.0.8
mypy 0.910
mypy-extensions 0.4.3
nbclient 0.5.4
nbconvert 6.1.0
nbformat 5.1.3
nest-asyncio 1.5.1
netron 7.6.6
notebook 6.4.3
numpy 1.24.3
oauthlib 3.2.2
opt-einsum 3.3.0
optax 0.1.8
orbax-checkpoint 0.2.3
packaging 23.2
pandas 2.0.3
pandocfilters 1.4.3
parso 0.8.2
partd 1.4.1
pbr 5.6.0
pep8 1.7.1
pexpect 4.8.0
pickleshare 0.7.5
pip 24.0
pipenv 2021.5.29
pkginfo 1.7.1
platformdirs 2.3.0
prometheus-client 0.11.0
prompt-toolkit 3.0.20
protobuf 4.25.3
ptyprocess 0.7.0
pyasn1 0.6.0
pyasn1_modules 0.4.0
pycodestyle 2.7.0
pycparser 2.20
pydocstyle 6.1.1
pyflakes 2.3.1
Pygments 2.18.0
pylama 7.7.1
pylint 2.10.2
pyparsing 2.4.7
pyrsistent 0.18.0
python-dateutil 2.8.2
pytz 2024.1
PyYAML 5.4.1
pyzmq 22.2.1
readme-renderer 29.0
requests 2.26.0
requests-oauthlib 2.0.0
requests-toolbelt 0.9.1
rfc3986 1.5.0
rich 13.7.1
rope 0.19.0
rsa 4.9
scipy 1.10.1
SecretStorage 3.3.1
Send2Trash 1.8.0
setuptools 58.0.4
six 1.16.0
smmap 4.0.0
snowballstemmer 2.1.0
stevedore 3.4.0
tensorboard 2.13.0
tensorboard-data-server 0.7.2
tensorflow 2.13.1
tensorflow-decision-forests 1.5.0
tensorflow-estimator 2.13.0
tensorflow-hub 0.16.1
tensorflow-io-gcs-filesystem 0.34.0
tensorflowjs 4.19.0
tensorstore 0.1.45
termcolor 2.4.0
terminado 0.12.1
testpath 0.5.0
tf-keras 2.15.0
toml 0.10.2
toolz 0.12.1
tornado 6.1
tqdm 4.62.2
traitlets 5.1.0
twine 3.4.2
typing_extensions 4.5.0
tzdata 2024.1
urllib3 1.26.6
virtualenv 20.7.2
virtualenv-clone 0.5.7
wcwidth 0.2.5
webencodings 0.5.1
Werkzeug 3.0.3
wheel 0.37.0
wrapt 1.12.1
wurlitzer 3.1.0
zipp 3.5.0
More attempts I think it is my python version is either to recent or not old enough
new instructions
python.exe -m pip install --upgrade pip
python -m venv myenv10 myenv10\scripts\activate
try the newest versions but if that doesn't work these versions will work. pip install tensorflow==2.13.1 pip install tensorflow-decision-forests==1.4.0 pip install tensorflowjs==4.19.0 --no-deps
pip install tensorflow-decision-forests==1.8.0 --no-deps pip install tensorflow-decision-forests==1.5.0
--ignore-installed tensorflow_decision_forests tensorflow tensorflow-io-gcs-filesystem tensorstore
tflite_convert --help tensorflowjs_converter --help
pip install tensorflowjs --no-deps
pip install tensorflowjs==4.19.0
pip install tensorflow==2.13.1 pip install tensorflowjs==4.19.0
tflite_convert --help tensorflowjs_converter --help
#!/bin/bash
tensorflowjs_converter --input_format=tfjs_layers_model --output_format=keras_saved_model ./model.json ./ tflite_convert --keras_model_file ./ --output_file ./model.tflite xxd -i model.tflite model.h
python.exe -m pip install --upgrade pip pip3 install --upgrade pip
python -m venv myenv2 myenv2\scripts\activate
pip3 install tensorflowjs pip3 install tensorflow==2.15.0 pip3 install tensorflow-hub pip3 install netron "dask[delayed]"
$env:TF_ENABLE_ONEDNN_OPTS=0
tflite_convert --help tensorflowjs_converter --help
xxd --help
https://sourceforge.net/projects/xxd-for-windows/
in power shell try
Format-Hex '.\your-file-name'