Layer names duplicated with multiple image inputs

[simoncozens]

Bug Description

When multiple ImageBlocks are used in the same model, the layer names they generate are not unique, so the model does not build.

Bug Reproduction

input_node1 = ak.ImageInput()
input_node2 = ak.ImageInput()
output_node1 = ak.ImageBlock(augment=False)(input_node1)
output_node2 = ak.ImageBlock(augment=False)(input_node2)
merged = ak.Merge()([output_node1, output_node2])
output_node = ak.ClassificationHead()(merged)
clf = ak.AutoModel(
    inputs=[input_node1, input_node2],
    outputs=output_node,
    overwrite=True,
    max_trials=5)

Expected Behavior

The model should build. Instead:

ValueError: The name "conv1_pad" is used 2 times in the model. All layer names should be unique.

Setup Details

Include the details about the versions of:

• OS type and version: Colab
• Python: 3.6
• autokeras: 1.0.3
• keras-tuner: 1.0.2rc0
• scikit-learn: 0.22.2.post1
• numpy: 1.18.5
• pandas: 1.0.5
• tensorflow: 2.2.0

[rakshith291]

Is there any update on this issue ? I'm facing the same problem as well

[Unemyr]

Also interested in this issue - and is it applicable to other block types as well?

[rakshith291]

Renaming the layers in the following way solved the problem for me.

for layer in model_con_1.layers : layer._name = layer.name + str('_1') for layer in model_con_2.layers : layer._name = layer.name + str('_2')

[vecrvecr]

Any update? I'm facing this problem trying to create an AutoModel with multiple ImageInputs

[abe-martin]

Bump on this one. Is it possible to use multiple image inputs with AutoModel? I see that rakshith291 posted some code to rename layers, but where does that go, somewhere in the source of the fit function?

[EmberaThomas]

same issue, I have tried to change autokeras _convert_to_dataset function but not work