C2W4- multi-classifier - Error during training

Course Q&A TensorFlow Developer Professional Convolutional Neural Networks in TensorFlow
1

course: “DeepLearning.AI TensorFlow Developer Professional Certificate” course.

I am facing error while training the model even after all the above code outputs are displayed as expected output.

  1. parse_data_from_input - this function working as expected
  2. train_val_generators - this function working as expected
  3. create_model - created two convolution layers as described and updated final layer as 26

1 frames
/usr/local/lib/python3.8/dist-packages/keras/engine/training.py in tf__train_function(iterator)
13 try:
14 do_return = True
—> 15 retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
16 except:
17 do_return = False

ValueError: in user code:

File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1051, in train_function  *
    return step_function(self, iterator)
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1040, in step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1030, in run_step  **
    outputs = model.train_step(data)
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 890, in train_step
    loss = self.compute_loss(x, y, y_pred, sample_weight)
File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 948, in compute_loss
    return self.compiled_loss(
File "/usr/local/lib/python3.8/dist-packages/keras/engine/compile_utils.py", line 201, in __call__
    loss_value = loss_obj(y_t, y_p, sample_weight=sw)
File "/usr/local/lib/python3.8/dist-packages/keras/losses.py", line 139, in __call__
    losses = call_fn(y_true, y_pred)
File "/usr/local/lib/python3.8/dist-packages/keras/losses.py", line 243, in call  **
    return ag_fn(y_true, y_pred, **self._fn_kwargs)
File "/usr/local/lib/python3.8/dist-packages/keras/losses.py", line 1787, in categorical_crossentropy
    return backend.categorical_crossentropy(
File "/usr/local/lib/python3.8/dist-packages/keras/backend.py", line 5119, in categorical_crossentropy
    target.shape.assert_is_compatible_with(output.shape)

ValueError: Shapes (None, 1) and (None, 26) are incompatible
2

I think the final layer should be one output instead of 26!

3

Actually it is a multi classifier with 26 letters of english alphabet. so, should be 26 not 1.

I see that if i give 1 , this error goes away. But it should be 26 only

4

Im not familiar with the assignment but if its asking for 26 outputs and if you are using binary crossentropy then you have a problem. For multiclass outputs you need categorical crossentropy.

5

I was using categorical crossentropy only but using sparse categorical crossentropy fixed the problem.

Thanks @gent.spah for looking into this. Appreciated!

6 Likes
6

had same issue, thanks for the solve. Turns out ‘categorical_crossentropy’ is for one-hot encoding (e.g. [1,0,0] for rock, [0,1,0] for paper, as was in the lesson), but ‘sparce_categorical_crossentropy’ is for ordinal encoding (numerical value for each category e.g. (0 for a , 1 for b, 2 for c) as in assignment)

4 Likes
7

Thank you for this info. Ran into the same Value error.
I think that the lectures should cover this essential piece of concept
and explain the differences between various loss functions!