Tensorflow -- Iterating over training and validation sequencially - python-3.x

Question

I have been going throught the Dataset API of tensorflow to feed different dataset with ease to an RNN model.
I got everything working following the not so many blogs together with the docs in the tensorflow website. My working example did the following:
--- Train on X epochs in a training dataset -> validate after all the training has concluded in a validation dataset.
However, I'm unable to develop the following example:
--- Train on X epochs in a training dataset -> validate in each epoch the training model with a validation dataset (a bit like what Keras does)
The problematic issue comes because of the following piece of code:
train_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE, drop_remainder=True).repeat()
val_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE_VAL, drop_remainder=True).repeat()
itr = tf.data.Iterator.from_structure(train_dataset.output_types, train_dataset.output_shapes)
train_init_op = itr.make_initializer(train_dataset)
validation_init_op = itr.make_initializer(val_dataset)
When I create the iterator from_structure, I need to specify an output_shape. Obviously, the output shape of the train dataset and the validation dataset is not the same as they have a different batch_size. However, the validation_init_op is throwing the following error, which it seems counterintuitive because validation sets have always different batch_size:
TypeError: Expected output shapes compatible with (TensorShape([Dimension(256), Dimension(11), Dimension(74)]), TensorShape([Dimension(256), Dimension(3)])) but got dataset with output shapes (TensorShape([Dimension(28), Dimension(11), Dimension(74)]), TensorShape([Dimension(28), Dimension(3)])).
I want to do this second approach to evaluate my model and see the common train and validation plots developed at the same time, to see how can I improve it (stopping the learning early and etc). However, with the first simple approach I don't get all this.
So, the question is: ¿Am I doing something wrong? ¿Does my second approach has to be tackled differently? I can think of creating two iterators, but I don't know if that is the right approach. Also, this answer by #MatthewScarpino points out to a feedable iterator because switching between reinitializable ones makes them to start all over again; however, the above error is not related with that part of the code -- ¿Maybe the reinitializable iterator is not intended to set a different batch size for the validation set and to only iterate it once after training whatever the size it is and without setting it in the .batch() method?
Any help is very much appreciated.
Full code for reference:
N_TIMESTEPS_X = xt.shape[0] ## The stack number
BATCH_SIZE = 256
#N_OBSERVATIONS = xt.shape[1]
N_FEATURES = xt.shape[2]
N_OUTPUTS = yt.shape[1]
N_NEURONS_LSTM = 128 ## Number of units in the LSTMCell
N_EPOCHS = 350
LEARNING_RATE = 0.001
### Define the placeholders anda gather the data.
xt = xt.transpose([1,0,2])
xval = xval.transpose([1,0,2])
train_data = (xt, yt)
validation_data = (xval, yval)
N_BATCHES = train_data[0].shape[0] // BATCH_SIZE
print('The number of batches is: {}'.format(N_BATCHES))
BATCH_SIZE_VAL = validation_data[0].shape[0] // N_BATCHES
print('The validation batch size is: {}'.format(BATCH_SIZE_VAL))
## We define the placeholders as a trick so that we do not break into memory problems, associated with feeding the data directly.
'''As an alternative, you can define the Dataset in terms of tf.placeholder() tensors, and feed the NumPy arrays when you initialize an Iterator over the dataset.'''
batch_size = tf.placeholder(tf.int64)
x = tf.placeholder(tf.float32, shape=[None, N_TIMESTEPS_X, N_FEATURES], name='XPlaceholder')
y = tf.placeholder(tf.float32, shape=[None, N_OUTPUTS], name='YPlaceholder')
# Creating the two different dataset objects.
train_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE, drop_remainder=True).repeat()
val_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE_VAL, drop_remainder=True).repeat()
# Creating the Iterator type that permits to switch between datasets.
itr = tf.data.Iterator.from_structure(train_dataset.output_types, train_dataset.output_shapes)
train_init_op = itr.make_initializer(train_dataset)
validation_init_op = itr.make_initializer(val_dataset)
next_features, next_labels = itr.get_next()

Answer 1

After investigating the best way to do this, I came across with this final implementation that works well on my end. Surely not be the best. So as to maintain the state, I used a feedable iterator.
AIM: This code is intented to be used when you want to train and validate at the same time, preserving the state of each iterator (i.e. validate with the newest model parameters). Together with that, the code saves the model and other stuff, like some information about the hyperparameters and summaries to visualize the training and validation in Tensorboard.
Also, don't get confused: you don't need to have a different batch size for the training set and for the validation set. This is a misconception that I have. The batch sizes must be the same AND you have to deal with the different number of batches, just passing when no more batches are left. This is a requirement so that you can create the iterator, regarding having both datasets the same data type and shape.
Hope it helps others. Just ignore the code that does not relate to your objectives. Many thanks for #kvish for all the help and time.
Code:
def RNNmodelTF(xt, yt, xval, yval, xtest, ytest):
N_TIMESTEPS_X = xt.shape[0] ## The stack number
BATCH_SIZE = 256
#N_OBSERVATIONS = xt.shape[1]
N_FEATURES = xt.shape[2]
N_OUTPUTS = yt.shape[1]
N_NEURONS_LSTM = 128 ## Number of units in the LSTMCell
N_EPOCHS = 350
LEARNING_RATE = 0.001
### Define the placeholders anda gather the data.
xt = xt.transpose([1,0,2])
xval = xval.transpose([1,0,2])
train_data = (xt, yt)
validation_data = (xval, yval)
N_BATCHES = train_data[0].shape[0] // BATCH_SIZE
## We define the placeholders as a trick so that we do not break into memory problems, associated with feeding the data directly.
'''As an alternative, you can define the Dataset in terms of tf.placeholder() tensors, and feed the NumPy arrays when you initialize an Iterator over the dataset.'''
batch_size = tf.placeholder(tf.int64)
x = tf.placeholder(tf.float32, shape=[None, N_TIMESTEPS_X, N_FEATURES], name='XPlaceholder')
y = tf.placeholder(tf.float32, shape=[None, N_OUTPUTS], name='YPlaceholder')
# Creating the two different dataset objects.
train_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE, drop_remainder=True).repeat()
val_dataset = tf.data.Dataset.from_tensor_slices((x,y)).batch(BATCH_SIZE, drop_remainder=True).repeat()
#################### Creating the Iterator type that permits to switch between datasets.
handle = tf.placeholder(tf.string, shape = [])
iterator = tf.data.Iterator.from_string_handle(handle, train_dataset.output_types, train_dataset.output_shapes)
next_features, next_labels = iterator.get_next()
train_val_iterator = tf.data.Iterator.from_structure(train_dataset.output_types, train_dataset.output_shapes)
train_iterator = train_val_iterator.make_initializer(train_dataset)
val_iterator = train_val_iterator.make_initializer(val_dataset)
###########################
### Create the graph
cellType = tf.nn.rnn_cell.LSTMCell(num_units=N_NEURONS_LSTM, name='LSTMCell')
inputs = tf.unstack(next_features, axis=1)
'''inputs: A length T list of inputs, each a Tensor of shape [batch_size, input_size]'''
RNNOutputs, _ = tf.nn.static_rnn(cell=cellType, inputs=inputs, dtype=tf.float32)
out_weights = tf.get_variable("out_weights", shape=[N_NEURONS_LSTM, N_OUTPUTS], dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
out_bias = tf.get_variable("out_bias", shape=[N_OUTPUTS], dtype=tf.float32, initializer=tf.zeros_initializer())
predictionsLayer = tf.matmul(RNNOutputs[-1], out_weights) + out_bias
### Define the cost function, that will be optimized by the optimizer.
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(logits=predictionsLayer, labels=next_labels, name='Softmax_plus_Cross_Entropy'))
optimizer_type = tf.train.AdamOptimizer(learning_rate=LEARNING_RATE, name='AdamOptimizer')
optimizer = optimizer_type.minimize(cost)
### Model evaluation
correctPrediction = tf.equal(tf.argmax(predictionsLayer,1), tf.argmax(next_labels,1))
accuracy = tf.reduce_mean(tf.cast(correctPrediction,tf.float32))
confusionMatrix1 = tf.confusion_matrix(tf.argmax(next_labels,1), tf.argmax(predictionsLayer,1), num_classes=3, name='ConfMatrix')
## Saving variables so that we can restore them afterwards.
saver = tf.train.Saver()
save_dir = '/media/SecondDiskHDD/8classModels/DLmodels/tfModels/{}_{}'.format(cellType.__class__.__name__, datetime.now().strftime("%Y%m%d%H%M%S"))
#save_dir = '/home/Desktop/tfModels/{}_{}'.format(cellType.__class__.__name__, datetime.now().strftime("%Y%m%d%H%M%S"))
os.mkdir(save_dir)
varDict = {'nTimeSteps': N_TIMESTEPS_X, 'BatchSize': BATCH_SIZE, 'nFeatures': N_FEATURES,
'nNeuronsLSTM': N_NEURONS_LSTM, 'nEpochs': N_EPOCHS,
'learningRate': LEARNING_RATE, 'optimizerType': optimizer_type.__class__.__name__}
varDicSavingTxt = save_dir + '/varDict.txt'
modelFilesDir = save_dir + '/modelFiles'
os.mkdir(modelFilesDir)
logDir = save_dir + '/TBoardLogs'
os.mkdir(logDir)
acc_summary = tf.summary.scalar('Accuracy', accuracy)
loss_summary = tf.summary.scalar('Cost_CrossEntropy', cost)
summary_merged = tf.summary.merge_all()
with open(varDicSavingTxt, 'w') as outfile:
outfile.write(repr(varDict))
with tf.Session() as sess:
tf.set_random_seed(2)
sess.run(tf.global_variables_initializer())
train_writer = tf.summary.FileWriter(logDir + '/train', sess.graph)
validation_writer = tf.summary.FileWriter(logDir + '/validation')
# initialise iterator with data
train_val_string = sess.run(train_val_iterator.string_handle())
cm1Total = None
cm2Total = None
print('¡Training starts!')
for epoch in range(N_EPOCHS):
batchAccList = []
batchAccListVal = []
tot_loss_train = 0
tot_loss_validation = 0
for batch in range(N_BATCHES):
sess.run(train_iterator, feed_dict = {x : train_data[0], y: train_data[1], batch_size: BATCH_SIZE})
optimizer_output, loss_value, summary, accBatch, cm1 = sess.run([optimizer, cost, summary_merged, accuracy, confusionMatrix1], feed_dict = {handle: train_val_string})
npArrayPred = predictionsLayer.eval(feed_dict= {handle: train_val_string})
predLabEnc = np.apply_along_axis(thresholdSet, 1, npArrayPred, value=0.5)
npArrayLab = next_labels.eval(feed_dict= {handle: train_val_string})
labLabEnc = np.argmax(npArrayLab, 1)
cm2 = confusion_matrix(labLabEnc, predLabEnc)
tot_loss_train += loss_value
batchAccList.append(accBatch)
try:
sess.run(val_iterator, feed_dict = {x: validation_data[0], y: validation_data[1], batch_size: BATCH_SIZE})
valLoss, valAcc, summary_val = sess.run([cost, accuracy, summary_merged], feed_dict = {handle: train_val_string})
tot_loss_validation += valLoss
batchAccListVal.append(valAcc)
except tf.errors.OutOfRangeError:
pass
if cm1Total is None and cm2Total is None:
cm1Total = cm1
cm2Total = cm2
else:
cm1Total += cm1
cm2Total += cm2
if batch % 10 == 0:
train_writer.add_summary(summary, batch)
validation_writer.add_summary(summary_val, batch)
epochAcc = tf.reduce_mean(batchAccList)
sess.run(train_iterator, feed_dict = {x : train_data[0], y: train_data[1], batch_size: BATCH_SIZE})
epochAcc_num = sess.run(epochAcc, feed_dict = {handle: train_val_string})
epochAccVal = tf.reduce_mean(batchAccListVal)
sess.run(val_iterator, feed_dict = {x: validation_data[0], y: validation_data[1], batch_size: BATCH_SIZE})
epochAcc_num_Val = sess.run(epochAccVal, feed_dict = {handle: train_val_string})
if epoch%10 == 0:
print("Epoch: {}, Loss: {:.4f}, Accuracy: {:.3f}".format(epoch, tot_loss_train / N_BATCHES, epochAcc_num))
print('Validation Loss: {:.4f}, Validation Accuracy: {:.3f}'.format(tot_loss_validation / N_BATCHES, epochAcc_num_Val))
cmLogFile1 = save_dir + '/cm1File.txt'
with open(cmLogFile1, 'w') as outfile:
outfile.write(repr(cm1Total))
cmLogFile2 = save_dir + '/cm2File.txt'
with open(cmLogFile2, 'w') as outfile:
outfile.write(repr(cm2Total))
saver.save(sess, modelFilesDir + '/model.ckpt')