CODEWITHSUNDEEP
tensorflowkeras
smk
smk

Reputation: 35

keras stops working on first epoch

I am running an image classification model. This is where I got stuck. Tried downgrading keras version to 1.0.2 and running the script again didn't work.

Jupyter notebook just keeps processing and doesn't run anything after the first epoch, running code on keras 1.2 with python 3.5

OUTPUT:

/anaconda/envs/py35/lib/python3.5/site-packages/ipykernel/__main__.py:19: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(101, activation="softmax", kernel_initializer="glorot_uniform", kernel_regularizer=<keras.reg...)`
/anaconda/envs/py35/lib/python3.5/site-packages/ipykernel/__main__.py:21: UserWarning: Update your `Model` call to the Keras 2 API: `Model(outputs=Tensor("de..., inputs=Tensor("in...)`
/anaconda/envs/py35/lib/python3.5/site-packages/ipykernel/__main__.py:44: UserWarning: The semantics of the Keras 2 argument `steps_per_epoch` is not the same as the Keras 1 argument `samples_per_epoch`. `steps_per_epoch` is the number of batches to draw from the generator at each epoch. Basically steps_per_epoch = samples_per_epoch/batch_size. Similarly `nb_val_samples`->`validation_steps` and `val_samples`->`steps` arguments have changed. Update your method calls accordingly.
/anaconda/envs/py35/lib/python3.5/site-packages/ipykernel/__main__.py:44: UserWarning: Update your `fit_generator` call to the Keras 2 API: `fit_generator(<image_gen..., verbose=2, epochs=32, validation_steps=25250, validation_data=<image_gen..., steps_per_epoch=1183, callbacks=[<keras.ca...)`
Epoch 1/32

INPUT:

%%time
from keras.models import Sequential, Model
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D, ZeroPadding2D, GlobalAveragePooling2D, AveragePooling2D
from keras.layers.normalization import BatchNormalization
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, CSVLogger, LearningRateScheduler, ReduceLROnPlateau
from keras.optimizers import SGD
from keras.regularizers import l2
import keras.backend as K
import math

K.clear_session()

base_model = InceptionV3(weights='imagenet', include_top=False, input_tensor=Input(shape=(299, 299, 3)))
x = base_model.output
x = AveragePooling2D(pool_size=(8, 8))(x)
x = Dropout(.4)(x)
x = Flatten()(x)
predictions = Dense(n_classes, kernel_initializer='glorot_uniform', W_regularizer=l2(.0005), activation='softmax')(x)





model = Model(input=base_model.input, output=predictions)

opt = SGD(lr=.01, momentum=.9)
model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])

checkpointer = ModelCheckpoint(filepath='model4.{epoch:02d}-{val_loss:.2f}.hdf5', verbose=1, save_best_only=True)
csv_logger = CSVLogger('model4.log')

def schedule(epoch):
    if epoch < 15:
        return .01
    elif epoch < 28:
        return .002
    else:
        return .0004
lr_scheduler = LearningRateScheduler(schedule)

model.fit_generator(train_generator,
                    validation_data=test_generator,
                    nb_val_samples=X_test.shape[0],
                    samples_per_epoch=X_train.shape[0],
                    nb_epoch=32,
                    verbose=2,
                    callbacks=[lr_scheduler, csv_logger, checkpointer])

Upvotes: 2

Views: 8287

Answers (2)

hafiz ahmad hassan
hafiz ahmad hassan

Reputation: 124

Co-Lab provides limited memory upto(12 GB) in cloud which creates many issues while solving a problem. That's why only 300 images are used to train and test.when images was preprocessed with dimension 600x600 and batch size was set to 128 it Keras model freezed during epoch 1 .Compiler did not show this error.Actually the error was runtime limited memory which was unable to handle by CoLab because it gave only 12GB limited memory for usage. Solution to above mentioned problem was solved by changing batch size to 4 and reduce image dimension to 300x300 because with 600x600 it still not work. Conclusively,Recommend Solution is Make Images dimension and Batch_size small until you get no error. Run Again and Again by further changing batch size and image dimension small until there will no run time errors.

Upvotes: 0

Dr. Snoopy
Dr. Snoopy

Reputation: 56397

Try with verbose = 1 in your model.fit call, it will print the progress bar. It is probably working, but due to the value of 2 given to the verbose parameter, it will only print one line of output AFTER the epoch has ended, which might take some time depending on your CPU/GPU and quantity of data.

Upvotes: 4

Related Questions