Inputs has to be named after the first entry of the dict for models with multiple inputs

[henrysky]

Keras version: "3.4.1"

For model with multiple inputs, inputs has to be named after the first entry of the dictionary which I think is a bug.

Lets say I have a Keras model with two inputs named a_name and other_name. Error will be raise if I try to train such model with those named inputs

model.fit({"a_name": x_train_1, "other_name": x_train_2}, y_train)

but would be fine if their names are a_name and a_name_whatever or other_name and other_name_whatever

Here is a minimal example on Google Colab to demostrate the issue (https://colab.research.google.com/drive/1oJ5QGMYV7GlHimkquu1SJWd4vyn9-ok0?usp=sharing).

All cells work without raising error if I use Keras 3.3.3

[Grvzard]

model = keras.Model(inputs=[x, x2], outputs=output)

=> model = keras.Model(inputs={input1_name: x, input2_name: x2}, outputs=output)

[mehtamansi29]

Hi @henrysky -

Keras 3.4.1 doesn't directly support dictionary input with model.fit, if the model has named inputs. More details on this you can find here.

def test_case(input1_name, input2_name):
  x = keras.Input(shape=[28, 28, 1], name=input1_name)
  x2 = keras.Input(shape=[10, ], name=input2_name)
  cnn_layer_1 = keras.layers.Conv2D(filters=2,kernel_size=[2,2],)(x)
  maxpool_1 = keras.layers.MaxPooling2D(pool_size=(2, 2))(cnn_layer_1)
  flatten = keras.layers.Flatten()(cnn_layer_1)
  flatten_concat = keras.layers.concatenate([flatten, x2])
  output = keras.layers.Dense(10, activation="softmax")(flatten)

  model = keras.Model(inputs={input1_name: x, input2_name: x2}, outputs=output)
  model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])
  model.fit({input1_name: x_train, input2_name: y_train}, y_train)

[github-actions]

This issue is stale because it has been open for 14 days with no activity. It will be closed if no further activity occurs. Thank you.

[mehtamansi29]

Hi @henrysky -

Does the issue is still reproduce to you ?

[henrysky]

The example by you has fixed the issue for me, thanks!

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No