Error: Unknown layer: Rescaling.

[co2nut]

Getting the error when trying to load custom converted model :
Error: Unknown layer: Rescaling. This may be due to one of the following reasons:

1. The layer is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
2. The custom layer is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().

[rthadur]

@co2nut this is a feature request , would you be willing to contribute ?

[roca77]

Have you guys found a solution to this? I am getting the same error...

[chihuahua]

Same, I am seeing this when loading a Keras model, ie

  tf.loadLayersModel('./model_export/model.json').then(
      (m: tf.GraphModel) => {
        this.model = m;
        console.log(m);
      });

[chihuahua]

My guess is that you all are using the tf.keras.layers.experimental.preprocessing.Rescaling layer, which has not yet been ported to TensorFlow JS.

[roca77]

My guess is that you all are using the tf.keras.layers.experimental.preprocessing.Rescaling layer, which has not yet been ported to TensorFlow JS.

Yes, Rescaling is causing the issue; I removed it and now it's working

[rthadur]

Thank you , closing this issue for now

[nwatab]

I know this issue is closed, but for visitors who uses tf.keras.applications.EfficientNetB?, it cannot be converted to tfjs because it uses Rescaling in it.
https://github.com/tensorflow/tensorflow/blob/v2.3.1/tensorflow/python/keras/applications/efficientnet.py#L316

[cardstdani]

I fixed it by adding this parameter at the conversion command: --output_format=tfjs_graph_model

After that, I loaded the model as Graph in the JavaScript code of the website. I hope it's useful

[narendraadloid]

My guess is that you all are using the tf.keras.layers.experimental.preprocessing.Rescaling layer, which has not yet been ported to TensorFlow JS.

Yes, Rescaling is causing the issue; I removed it and now it's working

I removed the scaling from the model but still it's not working
the model is defined below:

input_shape = (config.HEIGHT, config.WIDTH, 3)
base_model = keras.applications.MobileNetV3Large(weights='imagenet', input_shape=input_shape, include_top=False)
base_model.trainable = True
inputs = keras.Input(shape=input_shape)
x = base_model(inputs, training=True)
x = keras.layers.GlobalAveragePooling2D()(x)
x = keras.layers.Dropout(0.5)(x)
x = keras.layers.Dense(1024)(x)
x = keras.layers.Dense(256)(x)
x = keras.layers.Dense(64)(x)
x = keras.layers.Dense(8)(x)

outputs = keras.layers.Dense(2)(x)
model = keras.Model(inputs, outputs)
model.summary()