Tensorflow - Training error at the end of each epoch in progbar.py (keras)

TensorFlow
,
1

After changing laptops, the training of my transformer (a simple translator) fails at the end of the first epoch.

The code is pretty long but I use model.fit() and overridden train_step().

h = model.fit(
    ds.ds_tra,
    epochs=epochs,
    validation_data=ds.ds_val
)

The error is :

15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - masked_acc: 0.2338 - loss: 7.9628Traceback (most recent call last):
  File "C:\_P\dev\ai\AI1\tfe\translators\seq2seq_11\seq2seq_tst_train1.py", line 113, in <module>
    h = model.fit(
        ^^^^^^^^^^
  File "C:\Users\u2gil\AppData\Local\Programs\Python\Python311\Lib\site-packages\keras\src\utils\traceback_utils.py", line 122, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "C:\Users\u2gil\AppData\Local\Programs\Python\Python311\Lib\site-packages\keras\src\utils\progbar.py", line 92, in update
    self._values[k] = [v * value_base, value_base]
                       ~~^~~~~~~~~~~~
TypeError: unsupported operand type(s) for *: 'dict' and 'int'

The strange thing is that when I execute exactly the same code on my previous Windows10 laptop , it works fine.

The config of my new PC is :
core ultra 7 on Windows 11
Python 3.11.9
keras 3.6.0
numpy 1.26.4
tensorflow 2.17.0

2

Hi @Gilles_Jack, I guess this error is not related to Tensorflow and is a Python related error which occurs when you try to multiply integer and dictionary data types

integer=1
dic = {'key':'value'}
total = integer * dic
#OUTPUT: TypeError: unsupported operand type(s) for *: 'int' and 'dict'

Could you please check the data types of v and value_base variables in your code. Thank You.