Coremltools: TypeError: Keras layer of type <class 'tensorflow.python.keras.engine.functional.Functional'> is not supported.

[terd-dd]

## 🐞Describe the bug

my net:

Model: "functional_1"
Layer (type) Output Shape Param # Connected to
input_1 (InputLayer) [(None, 128, 128, 3) 0

---

block1_conv1 (Conv2D) (None, 128, 128, 64) 1792 input_1[0][0]

---

block1_conv2 (Conv2D) (None, 128, 128, 64) 36928 block1_conv1[0][0]

---

block1_pool (MaxPooling2D) (None, 64, 64, 64) 0 block1_conv2[0][0]

---

block2_conv1 (Conv2D) (None, 64, 64, 128) 73856 block1_pool[0][0]

---

block2_conv2 (Conv2D) (None, 64, 64, 128) 147584 block2_conv1[0][0]

---

block2_pool (MaxPooling2D) (None, 32, 32, 128) 0 block2_conv2[0][0]

---

block3_conv1 (Conv2D) (None, 32, 32, 256) 295168 block2_pool[0][0]

---

block3_conv2 (Conv2D) (None, 32, 32, 256) 590080 block3_conv1[0][0]

---

block3_conv3 (Conv2D) (None, 32, 32, 256) 590080 block3_conv2[0][0]

---

block3_pool (MaxPooling2D) (None, 16, 16, 256) 0 block3_conv3[0][0]

---

block4_conv1 (Conv2D) (None, 16, 16, 512) 1180160 block3_pool[0][0]

---

block4_conv2 (Conv2D) (None, 16, 16, 512) 2359808 block4_conv1[0][0]

---

block4_conv3 (Conv2D) (None, 16, 16, 512) 2359808 block4_conv2[0][0]

---

block4_pool (MaxPooling2D) (None, 8, 8, 512) 0 block4_conv3[0][0]

---

block5_conv1 (Conv2D) (None, 8, 8, 512) 2359808 block4_pool[0][0]

---

block5_conv2 (Conv2D) (None, 8, 8, 512) 2359808 block5_conv1[0][0]

---

block5_conv3 (Conv2D) (None, 8, 8, 512) 2359808 block5_conv2[0][0]

---

block5_pool (MaxPooling2D) (None, 4, 4, 512) 0 block5_conv3[0][0]

---

flatten (Flatten) (None, 8192) 0 block5_pool[0][0]

---

dense (Dense) (None, 1024) 8389632 flatten[0][0]

---

dropout (Dropout) (None, 1024) 0 dense[0][0]

---

up_sampling2d (UpSampling2D) (None, 12, 12, 512) 0 block5_pool[0][0]

---

dense_1 (Dense) (None, 1024) 1049600 dropout[0][0]

---

activation (Activation) (None, 12, 12, 512) 0 up_sampling2d[0][0]

---

dropout_1 (Dropout) (None, 1024) 0 dense_1[0][0]

---

conv2d (Conv2D) (None, 10, 10, 1) 4609 activation[0][0]

---

probabilistic_output (Dense) (None, 5) 5125 dropout_1[0][0]

---

positional_output (Reshape) (None, 10, 10) 0 conv2d[0][0]

Total params: 24,163,654
Trainable params: 24,163,654
Non-trainable params: 0

---

Trace

TypeError: Keras layer of type is not supported.

To Reproduce

coreml_model = coremltools.converters.keras.convert('model.h5', input_names=['image'], output_names=['output'],
class_labels=output_labels,
image_input_names='image')

System environment (please complete the following information):

coremltools version 4.0b2
macOS version- 10.15.16
virtualenv
python version 3.8
TensorFlow version 2.3.0
Keras version 2.4.3

[nikitaame]

I'm also facing this issue. Any updates? You can reproduce this issue by trying to convert https://github.com/notAI-tech/fastPunct

Simply add:
self.model = coremltools.converters.keras.convert(self.model, input_names=['text'], output_names=['output']) self.model.save('punctuation.mlmodel')
at line 173 of fastpunc.py

[WestyHopkins]

I’m also having this issue!

[hosituanit]

I also have it too.

[pc11000]

same issue here

[dhrebeniuk]

This issue stops me from implementing a great idea.

[nikitaame]

Can anyone from Apple speak to when this could be investigated in order to support Keras' Functional feature?

[aseemw]

Since this is using TF 2.x can you try using the Unified Conversion API , ct.convert,
instead of the multi-backend Keras converter , ct.converters.keras.convert.

That is,

tf_keras_model = tf.keras.models.load_model("keras_model.h5")
mlmodel = ct.convert(tf_keras_model)

[dhrebeniuk]

Since this is using TF 2.x can you try using the Unified Conversion API , ct.convert,
instead of the multi-backend Keras converter , ct.converters.keras.convert.

That is,

tf_keras_model = tf.keras.models.load_model("keras_model.h5")
mlmodel = ct.convert(tf_keras_model)

Thanks, that's help in my case

[nikitaame]

Thanks for the tip @aseemw. That fixed this issue but now facing another in as mentioned here 😅. We could probably close this one out for now then since this worked for both @dhrebeniuk and me!

[pc11000]

Since this is using TF 2.x can you try using the Unified Conversion API , ct.convert,
instead of the multi-backend Keras converter , ct.converters.keras.convert.

That is,

tf_keras_model = tf.keras.models.load_model("keras_model.h5")
mlmodel = ct.convert(tf_keras_model)

which coremodeltools version are you using?
I get the error:

AttributeError: module 'coremltools' has no attribute 'convert'

[nikitaame]

@pc11000 convert was released in 4.0.2b https://github.com/apple/coremltools/releases/

[pc11000]

ok now I get :

TypeError: can't convert expression to float

[nikitaame]

@pc11000 please search StackOverflow or this GitHub repo for other people facing your new issue. Otherwise, open a new bug report with the required info. This thread is specifically for the issue TypeError: Keras layer of type <class 'tensorflow.python.keras.engine.functional.Functional'> is not supported.

[neermesh]

same issue here