Ytukyg

Why is Keras telling me weights and batch sizes are different? How can I fix this? (adding next(...) does not help here).
Thanks in advance; there's something I'm just not getting here.

Error is on evaluate_generator(): TypeError: Axis must be specified when shapes of a and weights differ.

from sklearn.utils import shuffle as identical_shuffle



SAMPLES_PER_BATCH=1

BATCHES_PER_EPOCH=1

BATCHES_PER_VALIDATION=1

EPOCHS_PER_SIMULATION=2



def generate_training(batch_size=64):

    validation_length = (int(len(data.train)*0.25) // batch_size) * batch_size

    while True:

        for i in range(validation_length,len(data.train),batch_size):

            x,y,n = identical_shuffle(data.train[i:i+batch_size],data.target[i:i+batch_size],data.context[i:i+batch_size])

            yield {'input':x,'target':n}, y



def generate_validation(batch_size=64):

    validation_length = (int(len(data.train)*0.25) // batch_size) * batch_size

    while True:

        for i in range(0,validation_length,batch_size):

            x,y,n = identical_shuffle(data.train[i:i+batch_size],data.target[i:i+batch_size],data.context[i:i+batch_size])

            yield {'input':x,'target':n}, y



for epoch in range(EPOCHS_PER_SIMULATION):

  for batch in range(BATCHES_PER_EPOCH):

    result_training = model.train_on_batch( *next(generate_training(batch_size=SAMPLES_PER_BATCH)) )

    # <redacted operations on result_training>

  result_validation = model.evaluate_generator( generate_validation(batch_size=SAMPLES_PER_BATCH), steps=BATCHES_PER_VALIDATION )

For comparison, the below runs okay with the same generator.

model.fit_generator(generator=generate_training(batch_size=SAMPLES_PER_BATCH),

                    validation_data=next(generate_validation(batch_size=SAMPLES_PER_BATCH)),

                    validation_steps=1,

                    steps_per_epoch=1,

                    epochs=1)

Full Traceback

TypeError                                 Traceback (most recent call last)

<ipython-input-37-962a23d537c3> in <module>()

     79     print(result_training)

     80   model.reset_states()

---> 81   result_validation = model.evaluate_generator( generate_validation(batch_size=SAMPLES_PER_BATCH), steps=BATCHES_PER_VALIDATION )

     82   #result_validation = model.test_on_batch( *next(generate_validation(batch_size=SAMPLES_PER_BATCH)) )

     83   print("VALIDATION OF EPOCH "+str(epoch))



/usr/local/lib/python3.6/dist-packages/keras/legacy/interfaces.py in wrapper(*args, **kwargs)

     89                 warnings.warn('Update your `' + object_name + '` call to the ' +

     90                               'Keras 2 API: ' + signature, stacklevel=2)

---> 91             return func(*args, **kwargs)

     92         wrapper._original_function = func

     93         return wrapper



/usr/local/lib/python3.6/dist-packages/keras/engine/training.py in evaluate_generator(self, generator, steps, max_queue_size, workers, use_multiprocessing, verbose)

   1470             workers=workers,

   1471             use_multiprocessing=use_multiprocessing,

-> 1472             verbose=verbose)

   1473 

   1474     @interfaces.legacy_generator_methods_support



/usr/local/lib/python3.6/dist-packages/keras/engine/training_generator.py in evaluate_generator(model, generator, steps, max_queue_size, workers, use_multiprocessing, verbose)

    375         if i not in stateful_metric_indices:

    376             averages.append(np.average([out[i] for out in outs_per_batch],

--> 377                                        weights=batch_sizes))

    378         else:

    379             averages.append(np.float64(outs_per_batch[-1][i]))



/usr/local/lib/python3.6/dist-packages/numpy/lib/function_base.py in average(a, axis, weights, returned)

   1140             if axis is None:

   1141                 raise TypeError(

-> 1142                     "Axis must be specified when shapes of a and weights "

   1143                     "differ.")

   1144             if wgt.ndim != 1:



TypeError: Axis must be specified when shapes of a and weights differ.