logits and labels must be broadcastable error in Tensorflow RNN

Question

I am new to Tensorflow and deep leaning. I am trying to see how the loss decreases over 10 epochs in my RNN model that I created to read a dataset from kaggle which contains credit card fraud data. I am trying to classify the transactions as fraud(1) and not fraud(0). When I try to run the below code I keep getting the below error:

> 2018-07-30 14:59:33.237749: W
> tensorflow/core/kernels/queue_base.cc:277]
> _1_shuffle_batch/random_shuffle_queue: Skipping cancelled enqueue attempt with queue not closed Traceback (most recent call last):  
> File
> "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1322, in _do_call
>     return fn(*args)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1307, in _run_fn
>     options, feed_dict, fetch_list, target_list, run_metadata)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1409, in _call_tf_sessionrun
>     run_metadata) tensorflow.python.framework.errors_impl.InvalidArgumentError: logits
> and labels must be broadcastable: logits_size=[1,2] labels_size=[1,24]
>    [[Node: softmax_cross_entropy_with_logits_sg =
> SoftmaxCrossEntropyWithLogits[T=DT_FLOAT,
> _device="/job:localhost/replica:0/task:0/device:CPU:0"](add, softmax_cross_entropy_with_logits_sg/Reshape_1)]]
> 
> During handling of the above exception, another exception occurred:
> 
> Traceback (most recent call last):   File
> "/home/suleka/Documents/untitled1/RNN_CrediCard.py", line 96, in
> <module>
>     train_neural_network(x)   File "/home/suleka/Documents/untitled1/RNN_CrediCard.py", line 79, in
> train_neural_network
>     _, c = sess.run([optimizer, cost], feed_dict={x: feature_batch, y: label_batch})   File
> "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 900, in run
>     run_metadata_ptr)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1135, in _run
>     feed_dict_tensor, options, run_metadata)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1316, in _do_run
>     run_metadata)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py",
> line 1335, in _do_call
>     raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.InvalidArgumentError: logits
> and labels must be broadcastable: logits_size=[1,2] labels_size=[1,24]
>    [[Node: softmax_cross_entropy_with_logits_sg =
> SoftmaxCrossEntropyWithLogits[T=DT_FLOAT,
> _device="/job:localhost/replica:0/task:0/device:CPU:0"](add, softmax_cross_entropy_with_logits_sg/Reshape_1)]]
> 
> Caused by op 'softmax_cross_entropy_with_logits_sg', defined at:  
> File "/home/suleka/Documents/untitled1/RNN_CrediCard.py", line 96, in
> <module>
>     train_neural_network(x)   File "/home/suleka/Documents/untitled1/RNN_CrediCard.py", line 63, in
> train_neural_network
>     cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,
> labels=y))   File
> "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/util/deprecation.py",
> line 250, in new_func
>     return func(*args, **kwargs)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/nn_ops.py",
> line 1968, in softmax_cross_entropy_with_logits
>     labels=labels, logits=logits, dim=dim, name=name)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/nn_ops.py",
> line 1879, in softmax_cross_entropy_with_logits_v2
>     precise_logits, labels, name=name)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/gen_nn_ops.py",
> line 7205, in softmax_cross_entropy_with_logits
>     name=name)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py",
> line 787, in _apply_op_helper
>     op_def=op_def)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/ops.py",
> line 3414, in create_op
>     op_def=op_def)   File "/home/suleka/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/ops.py",
> line 1740, in __init__
>     self._traceback = self._graph._extract_stack()  # pylint: disable=protected-access
> 
> InvalidArgumentError (see above for traceback): logits and labels must
> be broadcastable: logits_size=[1,2] labels_size=[1,24]     [[Node:
> softmax_cross_entropy_with_logits_sg =
> SoftmaxCrossEntropyWithLogits[T=DT_FLOAT,
> _device="/job:localhost/replica:0/task:0/device:CPU:0"](add, softmax_cross_entropy_with_logits_sg/Reshape_1)]]

Can anyone point out what I am doing wrong in my code and also any problem in my code if possible. Thank you in advance.

Shown below is my code:

import tensorflow as tf
from tensorflow.contrib import rnn



# cycles of feed forward and backprop
hm_epochs = 10
n_classes = 2
rnn_size = 128
col_size = 30
batch_size = 24
try_epochs = 1
fileName = "creditcard.csv"

def create_file_reader_ops(filename_queue):
    reader = tf.TextLineReader(skip_header_lines=1)
    _, csv_row = reader.read(filename_queue)
    record_defaults = [[1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1.], [1]]
    col1, col2, col3, col4, col5, col6, col7, col8, col9, col10, col11, col12, col13, col14, col15, col16, col17, col18, col19, col20, col21, col22, col23, col24, col25, col26, col27, col28, col29, col30, col31 = tf.decode_csv(csv_row, record_defaults=record_defaults)
    features = tf.stack([col1, col2, col3, col4, col5, col6, col7, col8, col9, col10, col11, col12, col13, col14, col15, col16, col17, col18, col19, col20, col21, col22, col23, col24, col25, col26, col27, col28, col29, col30])
    return features, col31


def input_pipeline(fName, batch_size, num_epochs=None):
    # this refers to multiple files, not line items within files
    filename_queue = tf.train.string_input_producer([fName], shuffle=True, num_epochs=num_epochs)
    features, label = create_file_reader_ops(filename_queue)
    min_after_dequeue = 10000 # min of where to start loading into memory
    capacity = min_after_dequeue + 3 * batch_size # max of how much to load into memory
    # this packs the above lines into a batch of size you specify:
    feature_batch, label_batch = tf.train.shuffle_batch(
        [features, label],
        batch_size=batch_size,
        capacity=capacity,
        min_after_dequeue=min_after_dequeue)
    return feature_batch, label_batch


creditCard_data, creditCard_label = input_pipeline(fileName, batch_size, try_epochs)


x = tf.placeholder('float',[None,col_size])
y = tf.placeholder('float')


def recurrent_neural_network_model(x):
    #giving the weights and biases random values
    layer ={ 'weights': tf.Variable(tf.random_normal([rnn_size, n_classes])),
            'bias': tf.Variable(tf.random_normal([n_classes]))}

    x = tf.split(x, 24, 0)
    print(x)

    lstm_cell = rnn.BasicLSTMCell(rnn_size)
    outputs, states = rnn.static_rnn(lstm_cell, x, dtype=tf.float32 )
    output = tf.matmul(outputs[-1], layer['weights']) + layer['bias']

    return output

def train_neural_network(x):
    prediction = recurrent_neural_network_model(x)
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))
    optimizer = tf.train.AdamOptimizer().minimize(cost)


    with tf.Session() as sess:

        gInit = tf.global_variables_initializer().run()
        lInit = tf.local_variables_initializer().run()
        coord = tf.train.Coordinator()
        threads = tf.train.start_queue_runners(coord=coord)
        for epoch in range(hm_epochs):
            epoch_loss = 0

            for counter in range(101):
                    feature_batch, label_batch = sess.run([creditCard_data, creditCard_label])
                    print(label_batch.shape)
                    _, c = sess.run([optimizer, cost], feed_dict={x: feature_batch, y: label_batch})
                    epoch_loss += c
            print('Epoch', epoch, 'compleated out of', hm_epochs, 'loss:', epoch_loss)



train_neural_network(x)

Answer 1

Make sure that the number of labels in the final classification layer is equal to the number of classes you have in your dataset. InvalidArgumentError (see above for traceback): logits and labels must be broadcastable: logits_size=[1,2] labels_size=[1,24] as shown in your question might suggest that you are have just two classes in your final classification layer while you actually need 24.

In my case, I had 7 classes in my dataset, but I mistakenly used 4 labels in the final classification layer. Therefore, I had to change from

tf.keras.layers.Dense(4, activation="softmax")

to

tf.keras.layers.Dense(7, activation="softmax")

Answer 2

When you say

    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))

the prediction and labels have incompatible shapes. You need to change how the predictions are computed to get one per example in your minibatch.

Answer 3

A little late to the party but I had the same error with a CNN, I messed around with different types of cross entropy and the error was resolved by using sparce_softmax_cross_entropy_with_logits().

cost = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits=prediction, labels=y))

Answer 4

This error occurs because there is a mismatch in the count of Predicted class and the input. I copied a code and encountered this error

This was the original code where output is 5 classes

model.add(Dense(5, activation = "softmax"))

In my case, I had 30 classes and correcting the classes count fixed it

model.add(Dense(30, activation = "softmax"))

Answer 5

I met the similar problems when using the CNN, the problem occurs because I changed the labels data type as np.uint8 in the Generator function while did nothing for the labels type in the rest of my code. I solved the problem by changing the labels type to uint8 in all of my code.

Answer 6

This happened to me when using Tensorflow DataGenerator with small batch size. Try increasing your batch size.

Answer 7

Check the input shape in the first convolutional layer and the input shape fed into the network using data.shape. There is high chance you have a different shape fed into your CNN.