Unknown initializer: GlorotUniform when loading Keras model

[GiantTreeLP]

TensorFlow.js version

Version 0.13.2

Browser version

Google Chrome Version 69.0.3497.100 (64-bit)

Describe the problem or feature request

I am currently trying to load a convolutional neural network model from Keras (version 2.2.4) that has been converted using the tensorflow_converter utility (pip tensorflowjs version 0.6.4).

Upon loading the model, I am met with the following error:

Uncaught (in promise) Error: Unknown initializer: GlorotUniform. This may be due to one of the following reasons:
1. The initializer is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
2. The custom initializer is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().
    at new t (tfjs@0.13.2:2)
    at deserializeKerasObject (tfjs@0.13.2:2)
    at deserializeInitializer (tfjs@0.13.2:2)
    at getInitializer (tfjs@0.13.2:2)
    at t [as constructor] (tfjs@0.13.2:2)
    at new t (tfjs@0.13.2:2)
    at e.fromConfig (tfjs@0.13.2:2)
    at deserializeKerasObject (tfjs@0.13.2:2)
    at deserialize (tfjs@0.13.2:2)
    at i (tfjs@0.13.2:2)

However, the GlorotUniform initializer is part of tensorflowjs (https://js.tensorflow.org/api/0.13.0/#initializers.glorotUniform).

I have attached the model.json file that I'd like to load to this issue: model.zip

Code to reproduce the bug / link to feature request

const model = tf.loadModel("./model.json");

[caisq]

Thanks, @GiantTreeLP, for reporting this. This is a bug and we will fix it soon.

[Soufiane-Fartit]

I had the problem using this command: loaded_model = model_from_json(loaded_model_json)

but when I add (tf.keras.models.): loaded_model = tf.keras.models.model_from_json(loaded_model_json)

it works just fine

[aditya1702]

@Soufiane-Fartit I am still getting this error in tensorflow-js. Do you how to solve this?

[Soufiane-Fartit]

@aditya1702 I'm sorry to disappoint, but I'm still a beginner and don't know much

[caisq]

@aditya1702 Can you share the following info 1) The version of keras or tensorflow that you used to generate the saved model from the Python side (I assume you've exported model from Python) 2) The version of TensorFlow.js you are using 3) The exact error message you are getting and the line of JavaScript code executing when that error happens 4) Ideally, share the model JSON (before and after conversion) with us if possible.

[AbdurRafae]

@Soufiane-Fartit I am still getting this error in tensorflow-js. Do you how to solve this?

Try adding the following in the model load function call.

custom_objects={"GlorotUniform": tf.keras.initializers.glorot_uniform}

[abhirupa-tech]

import keras new_model = keras.models.load_model('weights/relu_tensor1')

ValueError: Unknown initializer: GlorotUniform

My packages are:

packages in environment at /home/abhirupa/anaconda3/envs/keraenv:

#

Name Version Build Channel

_libgcc_mutex 0.1 main
absl-py 0.9.0 pypi_0 pypi astor 0.8.1 pypi_0 pypi attrs 19.3.0 pypi_0 pypi backcall 0.1.0 pypi_0 pypi bleach 3.1.5 pypi_0 pypi ca-certificates 2020.1.1 0
certifi 2018.8.24 py35_1
decorator 4.4.2 pypi_0 pypi defusedxml 0.6.0 pypi_0 pypi entrypoints 0.3 pypi_0 pypi gast 0.3.3 pypi_0 pypi google-pasta 0.2.0 pypi_0 pypi grpcio 1.29.0 pypi_0 pypi h5py 2.10.0 pypi_0 pypi importlib-metadata 1.6.0 pypi_0 pypi ipykernel 5.3.0 pypi_0 pypi ipython 7.9.0 pypi_0 pypi ipython-genutils 0.2.0 pypi_0 pypi ipywidgets 7.5.1 pypi_0 pypi jedi 0.17.0 pypi_0 pypi jinja2 2.11.2 pypi_0 pypi jsonschema 3.2.0 pypi_0 pypi jupyter 1.0.0 pypi_0 pypi jupyter-client 6.1.3 pypi_0 pypi jupyter-console 6.1.0 pypi_0 pypi jupyter-core 4.6.3 pypi_0 pypi keras 2.2.4 pypi_0 pypi keras-applications 1.0.8 pypi_0 pypi keras-preprocessing 1.1.2 pypi_0 pypi libedit 3.1.20181209 hc058e9b_0
libffi 3.2.1 hd88cf55_4
libgcc-ng 9.1.0 hdf63c60_0
libstdcxx-ng 9.1.0 hdf63c60_0
markdown 3.2.2 pypi_0 pypi markupsafe 1.1.1 pypi_0 pypi mistune 0.8.4 pypi_0 pypi nbconvert 5.6.1 pypi_0 pypi nbformat 5.0.6 pypi_0 pypi ncurses 6.2 he6710b0_1
notebook 6.0.3 pypi_0 pypi numpy 1.18.4 pypi_0 pypi openssl 1.0.2u h7b6447c_0
packaging 20.4 pypi_0 pypi pandocfilters 1.4.2 pypi_0 pypi parso 0.7.0 pypi_0 pypi pexpect 4.8.0 pypi_0 pypi pickleshare 0.7.5 pypi_0 pypi pip 10.0.1 py35_0
prometheus-client 0.8.0 pypi_0 pypi prompt-toolkit 2.0.10 pypi_0 pypi protobuf 3.12.2 pypi_0 pypi ptyprocess 0.6.0 pypi_0 pypi pygments 2.6.1 pypi_0 pypi pyparsing 2.4.7 pypi_0 pypi pyrsistent 0.16.0 pypi_0 pypi python 3.5.6 hc3d631a_0
python-dateutil 2.8.1 pypi_0 pypi pyyaml 5.3.1 pypi_0 pypi pyzmq 19.0.1 pypi_0 pypi qtconsole 4.7.4 pypi_0 pypi qtpy 1.9.0 pypi_0 pypi readline 7.0 h7b6447c_5
scipy 1.4.1 pypi_0 pypi send2trash 1.5.0 pypi_0 pypi setuptools 40.2.0 py35_0
six 1.15.0 pypi_0 pypi sqlite 3.31.1 h62c20be_1
tensorboard 1.14.0 pypi_0 pypi tensorflow 1.14.0 pypi_0 pypi tensorflow-estimator 1.14.0 pypi_0 pypi termcolor 1.1.0 pypi_0 pypi terminado 0.8.3 pypi_0 pypi testpath 0.4.4 pypi_0 pypi tk 8.6.8 hbc83047_0
tornado 6.0.4 pypi_0 pypi traitlets 4.3.3 pypi_0 pypi wcwidth 0.1.9 pypi_0 pypi webencodings 0.5.1 pypi_0 pypi werkzeug 1.0.1 pypi_0 pypi wheel 0.31.1 py35_0
widgetsnbextension 3.5.1 pypi_0 pypi wrapt 1.12.1 pypi_0 pypi xz 5.2.5 h7b6447c_0
zipp 1.2.0 pypi_0 pypi zlib 1.2.11 h7b6447c_3

[samyak1999]

@DeveloperAbhirupa Use the below codes. It works fine

from keras.models import load_model classifier = tf.keras.models.load_model("C:/Users/8_mnist_simple_cnn_10_Epochs.h5")