I am training a GRU neural network and added dropout and recurrent dropout in my GRU layer but since then I can't get reproducible results every time I run the program again and I can't fix this problem even with :
recurrent_initializer=tf.keras.initializers.Orthogonal(seed=42),
kernel_initializer=tf.keras.initializers.GlorotUniform(seed=42))
in the same layer.
This is my model:
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.GRU(20, activation='tanh',dropout=0.1,
recurrent_dropout=0.2,recurrent_activation="sigmoid", return_sequences =
False, input_shape=(train_XX.shape[1], train_XX.shape[2]),
recurrent_initializer=tf.keras.initializers.Orthogonal(seed=42),
kernel_initializer=tf.keras.initializers.GlorotUniform(seed=42)))
model.add(tf.keras.layers.Dense(1, activation='sigmoid',
kernel_initializer=tf.keras.initializers.GlorotUniform(seed=42),))
model.compile(loss=tf.keras.losses.BinaryCrossentropy(from_logits=False,
name="binary_crossentropy",),optimizer='adam',
metrics=[tf.keras.metrics.PrecisionAtRecall(0.75)] )
I had already set the seed at the beginning of the programme with:
import numpy as np
import tensorflow as tf
import random as rn
np.random.seed(1)
tf.random.set_seed(2)
rn.seed(3)
but by adding before the 3 rows of seed fixation:
import os
os.environ['PYTHONHASHSEED'] = '0'
os.environ['CUDA_VISIBLE_DEVICES'] = ''
it resolves my problem.
Related
My data set is satellite observation which includes a lot of zeroes so that highly effect my final simulation results.
I have two sets of input data, dynamic ones (X_dynamic_LSTM.shape (95931, 1, 5)) which change through time series and static ones (X_static_MLP.shape (95931, 10)) which is not change. For dynamic ones I used LSTM and for static ones the MLP. I Concatenate the two and get the final results by another MLP.
Can you suggest how should I ignore these zero variables in my prediction dataframe??? I know about Masking and Embedding but don't know how to add them in my code!
from tensorflow.keras.layers import Input, LSTM, Dense, Concatenate
from tensorflow.keras.models import Model
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Masking
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Embedding
lstm_input = Input(shape=(X_dynamic_LSTM.shape[1], X_dynamic_LSTM.shape[2]))
x = Masking(mask_value=0.)(lstm_input)
x = LSTM(70, activation='tanh', return_sequences=True)(x)
x = Dropout(0.3)(x)
x = LSTM(35)(x)
x = Dropout(0.3)(x)
x = Dense(1, activation='tanh')(x)
#mlp input with additonal 3 variables at t=t
mlp_input=Input(shape=(X_static_MLP.shape[1]))
mlp = Dense(30, activation='relu')(mlp_input)
mlp = Dense(20, activation='relu')(mlp)
merge = Concatenate()([x, mlp])
hidden1 = Dense(5, activation='relu')(merge)
mlp_out = Dense(1, activation='relu')(hidden1)
model = Model(inputs=[lstm_input, mlp_input],outputs=mlp_out)
#compile the model
model.compile(loss='mae', optimizer='adam')
#fit the model
model.fit([X_dynamic_LSTM, X_static_MLP], y_train, batch_size=40,
epochs=10, validation_split=0.2)
use embedding layer in your first layer
you can use this link
>>> model = tf.keras.Sequential()
>>> model.add(tf.keras.layers.Embedding())
I'm trying to do transfer learning with MobileNetV3 in Keras but I'm having some issues.
from keras.models import Model
from keras.layers import GlobalMaxPooling2D, Dense, Dropout
from keras.optimizers import Adam
from keras.callbacks import ModelCheckpoint
from tensorflow.keras.applications import MobileNetV3Small
import numpy as np
from tqdm import tqdm
from PIL import Image, ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
pretrained_model = MobileNetV3Small(input_shape=(224,224,3),
weights="imagenet",
include_top=False)
# freeze all layers except the last one
for layer in pretrained_model.layers:
layer.trainable = False
pretrained_model.layers[-1].trainable = True
# combine the model with some extra layers for classification
last_output = pretrained_model.layers[-1].output
x = GlobalMaxPooling2D()(last_output)
x = Dense(128, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(1, activation='sigmoid')(x)
model = Model(pretrained_model.input, x)
I get this error when I try to make the Dense layer:
TypeError: Cannot convert a symbolic Keras input/output to a numpy array. This error may indicate that you're trying to pass a symbolic value to a NumPy call, which is not supported. Or, you may be trying to pass Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing Keras from automatically converting the API call to a lambda layer in the Functional Model.
but it's fixed by adding the following code snippet:
from tensorflow.python.framework.ops import disable_eager_execution
disable_eager_execution()
When I include the code fix above, I get this error when I call model.fit():
FailedPreconditionError: 2 root error(s) found.
(0) Failed precondition: Could not find variable Conv_1_2/kernel. This could mean that the variable has been deleted. In TF1, it can also mean the variable is uninitialized. Debug info: container=localhost, status=Not found: Resource localhost/Conv_1_2/kernel/N10tensorflow3VarE does not exist.
[[{{node Conv_1_2/Conv2D/ReadVariableOp}}]]
[[_arg_dense_12_target_0_1/_100]]
(1) Failed precondition: Could not find variable Conv_1_2/kernel. This could mean that the variable has been deleted. In TF1, it can also mean the variable is uninitialized. Debug info: container=localhost, status=Not found: Resource localhost/Conv_1_2/kernel/N10tensorflow3VarE does not exist.
[[{{node Conv_1_2/Conv2D/ReadVariableOp}}]]
0 successful operations.
0 derived errors ignored.
How can I fix these issues and train the model?
From comments
Don't mix tf.keras and standalone keras. They are not compatible. Only use one of them (paraphrased from Frightera)
Working code as shown below
from tensorflow.keras.models import Model
from tensorflow.keras.layers import GlobalMaxPooling2D, Dense, Dropout
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ModelCheckpoint
from tensorflow.keras.applications import MobileNetV3Small
import numpy as np
from tqdm import tqdm
from PIL import Image, ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
pretrained_model = MobileNetV3Small(input_shape=(224,224,3),
weights="imagenet",
include_top=False)
# freeze all layers except the last one
for layer in pretrained_model.layers:
layer.trainable = False
pretrained_model.layers[-1].trainable = True
# combine the model with some extra layers for classification
last_output = pretrained_model.layers[-1].output
x = GlobalMaxPooling2D()(last_output)
x = Dense(128, activation='relu')(x)
x = Dropout(0.5)(x)
x = Dense(1, activation='sigmoid')(x)
model = Model(pretrained_model.input, x)
I'm implementing SRGAN (and am not very experienced in this field), which uses a pre-trained VGG19 model to extract features. The following code was working fine on Keras 2.1.2 and tf 1.15.0 till yesterday. then it started throwing an "AttributeError: module 'keras.utils.generic_utils' has no attribute 'populate_dict_with_module_objects'" So i updated the keras version to 2.4.3 and tf to 2.5.0. but then its showing a "Input 0 of layer fc1 is incompatible with the layer: expected axis -1 of input shape to have value 25088 but received input with shape (None, 32768)" on the following line
features = vgg(input_layer)
But here the input has to be (256,256,3).
I had downgraded the keras and tf versions to the one I mentioned before to get rid of this error in the first place and it was working well till yesterday.
changing the input shape to (224,224,3) does not work. Any help in solving this error will be very appreciated.
import glob
import time
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
import keras
from keras.layers import Input
from keras.applications.vgg19 import VGG19
from keras.callbacks import TensorBoard
from keras.layers import BatchNormalization, Activation, LeakyReLU, Add, Dense,Flatten
from keras.layers.convolutional import Conv2D, UpSampling2D
from keras.models import Model
from keras.optimizers import Adam
from scipy.misc import imread, imresize
from PIL import Image
def build_vgg():
input_shape = (256, 256, 3)
vgg = VGG19(weights="imagenet")
vgg.outputs = [vgg.layers[9].output]
input_layer = Input(shape=input_shape)
features = vgg(input_layer)
model = Model(inputs=[input_layer], outputs=[features])
return model
vgg = build_vgg()
vgg.trainable = False
vgg.compile(loss='mse', optimizer=common_optimizer, metrics=['accuracy'])
# Build and compile the discriminator
discriminator = build_discriminator()
discriminator.compile(loss='mse', optimizer=common_optimizer, metrics=['accuracy'])
# Build the generator network
generator = build_generator()
The Error message
Im using google colab
Importing keras from tensorflow and setting include_top=False in
vgg = VGG19(weights="imagenet",include_top=False)
seems to work.
I made a script in tensorflow 2.x but I had to downconvert it to tensorflow 1.x (tested in 1.14 and 1.15). However, the tf1 version performs very differently (10% accuracy lower on the test set). See also the plot for train and validation performance (diagram is attached below).
Looking at the operations needed for the migration from tf1 to tf2 it seems that only the Adam learning rate may be a problem but I'm defining it explicitly tensorflow migration
I've reproduced the same behavior both locally on GPU and CPU and on colab. The keras used was the one built-in in tensorflow (tf.keras). I've used the following functions (both for train,validation and test), using a sparse categorization (integers):
train_datagen = tf.keras.preprocessing.image.ImageDataGenerator(
horizontal_flip=horizontal_flip,
#rescale=None, #not needed for resnet50
preprocessing_function=None,
validation_split=None)
train_dataset = train_datagen.flow_from_directory(
directory=train_dir,
target_size=image_size,
class_mode='sparse',
batch_size=batch_size,
shuffle=True)
And the model is a simple resnet50 with a new layer on top:
IMG_SHAPE = img_size+(3,)
inputs = Input(shape=IMG_SHAPE, name='image_input',dtype = tf.uint8)
x = tf.cast(inputs, tf.float32)
# not working in this version of keras. inserted in imageGenerator
x = preprocess_input_resnet50(x)
base_model = tf.keras.applications.ResNet50(
include_top=False,
input_shape = IMG_SHAPE,
pooling=None,
weights='imagenet')
# Freeze the pretrained weights
base_model.trainable = False
x=base_model(x)
# Rebuild top
x = GlobalAveragePooling2D(data_format='channels_last',name="avg_pool")(x)
top_dropout_rate = 0.2
x = Dropout(top_dropout_rate, name="top_dropout")(x)
outputs = Dense(num_classes,activation="softmax", name="pred_out")(x)
model = Model(inputs=inputs, outputs=outputs,name="ResNet50_comp")
optimizer = tf.keras.optimizers.Adam(lr=learning_rate)
model.compile(optimizer=optimizer,
loss="sparse_categorical_crossentropy",
metrics=['accuracy'])
And then I'm calling the fit function:
history = model.fit_generator(train_dataset,
steps_per_epoch=n_train_batches,
validation_data=validation_dataset,
validation_steps=n_val_batches,
epochs=initial_epochs,
verbose=1,
callbacks=[stopping])
I've reproduced the same behavior for example with the following full script (applied to my dataset and changed to adam and removed intermediate final dense layer):
deep learning sandbox
The easiest way to replicate this behavior was to enable or disable the following line on a tf2 environment with the same script and add the following line to it. However, I've tested also on tf1 environments (1.14 and 1.15):
tf.compat.v1.disable_v2_behavior()
Sadly I cannot provide the dataset.
Update 26/11/2020
For full reproducibility I've obtained a similar behaviour by means of the food101 (101 categories) dataset enabling tf1 behaviour with 'tf.compat.v1.disable_v2_behavior()'. The following is the script executed with tensorflow-gpu 2.2.0:
#%% ref https://medium.com/deeplearningsandbox/how-to-use-transfer-learning-and-fine-tuning-in-keras-and-tensorflow-to-build-an-image-recognition-94b0b02444f2
import os
import sys
import glob
import argparse
import matplotlib.pyplot as plt
import tensorflow as tf
# enable and disable this to obtain tf1 behaviour
tf.compat.v1.disable_v2_behavior()
from tensorflow.keras import __version__
from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D
from tensorflow.keras.optimizers import Adam
# since i'm using resnet50 weights from imagenet, i'm using food101 for
# similar but different categorization tasks
# pip install tensorflow-datasets if tensorflow_dataset not found
import tensorflow_datasets as tfds
(train_ds,validation_ds),info= tfds.load('food101', split=['train','validation'], shuffle_files=True, with_info=True)
assert isinstance(train_ds, tf.data.Dataset)
print(train_ds)
#%%
IM_WIDTH, IM_HEIGHT = 224, 224
NB_EPOCHS = 10
BAT_SIZE = 32
def get_nb_files(directory):
"""Get number of files by searching directory recursively"""
if not os.path.exists(directory):
return 0
cnt = 0
for r, dirs, files in os.walk(directory):
for dr in dirs:
cnt += len(glob.glob(os.path.join(r, dr + "/*")))
return cnt
def setup_to_transfer_learn(model, base_model):
"""Freeze all layers and compile the model"""
for layer in base_model.layers:
layer.trainable = False
model.compile(optimizer='rmsprop', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
def add_new_last_layer(base_model, nb_classes):
"""Add last layer to the convnet
Args:
base_model: keras model excluding top
nb_classes: # of classes
Returns:
new keras model with last layer
"""
x = base_model.output
x = GlobalAveragePooling2D()(x)
#x = Dense(FC_SIZE, activation='relu')(x) #new FC layer, random init
predictions = Dense(nb_classes, activation='softmax')(x) #new softmax layer
model = Model(inputs=base_model.input, outputs=predictions)
return model
def train(nb_epoch, batch_size):
"""Use transfer learning and fine-tuning to train a network on a new dataset"""
#nb_train_samples = train_ds.cardinality().numpy()
nb_train_samples=info.splits['train'].num_examples
nb_classes = info.features['label'].num_classes
classes_names = info.features['label'].names
#nb_val_samples = validation_ds.cardinality().numpy()
nb_val_samples = info.splits['validation'].num_examples
#nb_epoch = int(args.nb_epoch)
#batch_size = int(args.batch_size)
def preprocess(features):
#print(features['image'], features['label'])
image = tf.image.resize(features['image'], [224,224])
#image = tf.divide(image, 255)
#print(image)
# data augmentation
image=tf.image.random_flip_left_right(image)
image = preprocess_input(image)
label = features['label']
# for categorical crossentropy
#label = tf.one_hot(label,101,axis=-1)
#return image, tf.cast(label, tf.float32)
return image, label
#pre-processing the dataset to fit a specific image size and 2D labelling
train_generator = train_ds.map(preprocess).batch(batch_size).repeat()
validation_generator = validation_ds.map(preprocess).batch(batch_size).repeat()
#train_generator=train_ds
#validation_generator=validation_ds
#fig = tfds.show_examples(validation_generator, info)
# setup model
base_model = ResNet50(weights='imagenet', include_top=False) #include_top=False excludes final FC layer
model = add_new_last_layer(base_model, nb_classes)
# transfer learning
setup_to_transfer_learn(model, base_model)
history = model.fit(
train_generator,
epochs=nb_epoch,
steps_per_epoch=nb_train_samples//BAT_SIZE,
validation_data=validation_generator,
validation_steps=nb_val_samples//BAT_SIZE)
#class_weight='auto')
#execute
history = train(nb_epoch=NB_EPOCHS, batch_size=BAT_SIZE)
And the performance on food101 dataset:
update 27/11/2020
It's possible to see the discrepancy also in the way smaller oxford_flowers102 dataset:
(train_ds,validation_ds,test_ds),info= tfds.load('oxford_flowers102', split=['train','validation','test'], shuffle_files=True, with_info=True)
Nb: the above plot shows confidences given by running the same training multiple times and evaluatind mean and std to check for the effects on random weights initialization and data augmentation.
Moreover I've tried some hyperparameter tuning on tf2 resulting in the following picture:
changing optimizer (adam and rmsprop)
not applying horizontal flipping aumgentation
deactivating keras resnet50 preprocess_input
Thanks in advance for every suggestion. Here are the accuracy and validation performance on tf1 and tf2 on my dataset:
Update 14/12/2020
I'm sharing the colab for reproducibility on oxford_flowers at the clic of a button:
colab script
I came across something similar, when doing the opposite migration (from TF1+Keras to TF2).
Running this code below:
# using TF2
import numpy as np
from tensorflow.keras.applications.resnet50 import ResNet50
fe = ResNet50(include_top=False, pooling="avg")
out = fe.predict(np.ones((1,224,224,3))).flatten()
sum(out)
>>> 212.3205274187726
# using TF1+Keras
import numpy as np
from keras.applications.resnet50 import ResNet50
fe = ResNet50(include_top=False, pooling="avg")
out = fe.predict(np.ones((1,224,224,3))).flatten()
sum(out)
>>> 187.23898954353717
you can see the same model from the same library on different versions does not return the same value (using sum as a quick check-up). I found the answer to this mysterious behavior in this other SO answer: ResNet model in keras and tf.keras give different output for the same image
Another recommendation I'd give you is, try using pooling from inside applications.resnet50.ResNet50 class, instead of the additional layer in your function, for simplicity, and to remove possible problem-generators :)
tensorflow.keras api not working on while creating the layers reference, any other methods of creating layers reference?
code :
layer=keras.layers
Error message : NameError: name 'leyer' is not defined
Full code is pasted here...
import tensorflow as tf
from tensorflow import keras
import pandas as pd
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
from sklearn.preprocessing import LabelEncoder
import numpy as np
#makin seed values
seed=7
np.random.seed(seed)
#setting up the dataset for training
dataframe=pd.read_csv("../datasets/iris.csv",header=None)
data=dataframe.values
input_x = data[:,0:4]
true_y = data[:,4]
#Encoding the true_y data to one hot encoding
le=LabelEncoder()
le.fit(true_y)
y_encoded = le.transform(true_y)
y_encoded = keras.utils.to_categorical(y_encoded,num_classes=3)
# creating the model
def base_fun():
layer=keras.layers
model = keras.models.Sequential()
model.add(layer.Dense(4,input_dim=4,kernel_initializer='normal',activation='relu'))
model.add(leyer.Dense(3, kernel_initializer='normal', activation='relu'))
estimator=keras.wrappers.scikit_learn.KerasClassifier(build_fn=base_fun,epochs=20,batch_size=10)
kfold = KFold(n_splits=10, shuffle=True, random_state=seed)
result = cross_val_score(estimator, input_x, y_encoded,cv=kfold)
print("Accuracy : %.2%% (%.2%%)" %(result.mean()*100, result.std()*100))
Well, this line:
model.add(leyer.Dnese(3, kernel_initializer='normal', activation='relu'))
has two typos, namely leyer should be layer and Dnese should be Dense like
model.add(layer.Dense(3, kernel_initializer='normal', activation='relu'))
Based on your comment, this line also causes an error:
estimator = keras.wrappers.scikit_learn.KerasClassifier( build_fn = base_fun, epochs = 20, batch_size = 10 )
From the Keras Scikit documentation:
build_fn should construct, compile and return a Keras model, which will then be used to fit/predict.
But you function base_fun() does not return anything. Append this line at the end of base_fun():
return model
As per your comment, the last print line could be changed to this (I don't know the % formatting, I generally use the syntax below):
print( "Accuracy : {:.2%} ({:.2%})".format( result.mean(), result.std() ) )