"Working with RNNs" guide includes code that throws an exception #2142
Description
Issue Type
Documentation Bug
Source
source
Keras Version
3.10.0
Custom Code
No
OS Platform and Distribution
Windows 10
Python version
3.12.7
GPU model and memory
No response
Current Behavior?
The code at https://www.tensorflow.org/guide/keras/working_with_rnns#define_a_custom_cell_that_supports_nested_inputoutput (Definition and use of NestedCell class) throws an exception having to do with unexpected tuple shapes. I am not experienced enough with this to be able to tell for sure what the cause is, but I suspect it has to do with changes that have been made in how shapes are handled in keras since this guide was written. It's probably quite easy to fix the example to make it work on latest versions, but I wouldn't know it at this point in time.
I am only sending this to improve the documentation. I have gotten what I wanted from this guide and this is not a problem for me at the present time.
The exception:
---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
Cell In[24], line 18
15 input_1 = keras.Input((None, i1))
16 input_2 = keras.Input((None, i2, i3))
---> 18 outputs = rnn((input_1, input_2))
20 model = keras.models.Model([input_1, input_2], outputs)
22 model.compile(optimizer="adam", loss="mse", metrics=["accuracy"])
File ~\AppData\Roaming\Python\Python312\site-packages\keras\src\utils\traceback_utils.py:122, in filter_traceback.<locals>.error_handler(*args, **kwargs)
119 filtered_tb = _process_traceback_frames(e.__traceback__)
120 # To get the full stack trace, call:
121 # `keras.config.disable_traceback_filtering()`
--> 122 raise e.with_traceback(filtered_tb) from None
123 finally:
124 del filtered_tb
Cell In[23], line 14, in NestedCell.build(self, input_shapes)
11 def build(self, input_shapes):
12 # expect input_shape to contain 2 items, [(batch, i1), (batch, i2, i3)]
13 i1 = input_shapes[0][1]
---> 14 i2 = input_shapes[1][1]
15 i3 = input_shapes[1][2]
17 self.kernel_1 = self.add_weight(
18 shape=(i1, self.unit_1), initializer="uniform", name="kernel_1"
19 )
IndexError: tuple index out of range
Standalone code to reproduce the issue or tutorial link
https://www.tensorflow.org/guide/keras/working_with_rnns#define_a_custom_cell_that_supports_nested_inputoutput