We Keep Coding

How to implement LSTM in tensorflow v1 from pandas dataframe

I've tried following tutorials on implementing this but I keep getting dimension errors on the LSTM layer.
ValueError: Input 0 of layer LSTM is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [None, 2]
import random
import numpy as np
import tensorflow as tf
from tensorflow import feature_column as fc
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, DenseFeatures, Reshape
from sklearn.model_selection import train_test_split
def df_to_dataset(features, target, batch_size=32):
return tf.data.Dataset.from_tensor_slices((dict(features), target)).batch(batch_size)
# Reset randomization seeds
np.random.seed(0)
tf.random.set_random_seed(0)
random.seed(0)
# Assume 'frame' to be a dataframe with 3 columns: 'optimal_long_log_return', 'optimal_short_log_return' (independent variables) and 'equilibrium_log_return' (dependent variable)
X = frame[['optimal_long_log_return', 'optimal_short_log_return']][:-1]
Y = frame['equilibrium_log_return'].shift(-1)[:-1]
X_train, _X, y_train, _y = train_test_split(X, Y, test_size=0.5, shuffle=False, random_state=1)
X_validation, X_test, y_validation, y_test = train_test_split(_X, _y, test_size=0.5, shuffle=False, random_state=1)
train = df_to_dataset(X_train, y_train)
validation = df_to_dataset(X_validation, y_validation)
test = df_to_dataset(X_test, y_test)
feature_columns = [fc.numeric_column('optimal_long_log_return'), fc.numeric_column('optimal_short_log_return')]
model = Sequential()
model.add(DenseFeatures(feature_columns, name='Metadata'))
model.add(LSTM(256, name='LSTM'))
model.add(Dense(1, name='Output'))
model.compile(loss='logcosh', metrics=['mean_absolute_percentage_error'], optimizer='Adam')
model.fit(train, epochs=10, validation_data=validation, verbose=1)
loss, accuracy = model.evaluate(test, verbose=0)
print(f'Target Error: {accuracy}%')
After seeing this issue elsewhere I've tried setting input_shape=(None, *X_train.shape), input_shape=X_train.shape, neither works. I also tried inserting a Reshape layer model.add(Reshape(X_train.shape)) before the LSTM layer and it fixed the issue but I got another issue in its place:
InvalidArgumentError: Input to reshape is a tensor with 64 values, but the requested shape has 8000
...and I'm not even sure adding the Reshape layer is doing what I think it is doing. After all, why would reshaping the data to its own shape fix things? Something is happening with my data that I just don't understand.
Also, I'm using this for time series analysis (stock returns), so I would think that the LSTM model should be stateful and temporal. Would I need to move the timestamp index into its own column in the pandas database before converting to a tensor?
Unfortunately I'm obligated to use tensorflow v1.15 as this is being developed on the QuantConnect platform and they presumably won't be updating the library any time soon.
EDIT: I've made a bit of progress by using TimeseriesGenerator, but now I'm getting the following error (which returns no results on Google):
KeyError: 'No key found for either mapped or original key. Mapped Key: []; Original Key: []'
Code below (I'm sure I'm using the input_shape arguments incorrectly):
train = TimeseriesGenerator(X_train, y_train, 1, batch_size=batch_size)
validation = TimeseriesGenerator(X_validation, y_validation, 1, batch_size=batch_size)
test = TimeseriesGenerator(X_test, y_test, 1, batch_size=batch_size)
model = Sequential(name='Expected Equilibrium Log Return')
model.add(LSTM(256, name='LSTM', stateful=True, batch_input_shape=(1, batch_size, X_train.shape[1]), input_shape=(1, X_train.shape[1])))
model.add(Dense(1, name='Output'))
model.compile(loss='logcosh', metrics=['mean_absolute_percentage_error'], optimizer='Adam', sample_weight_mode='temporal')
print(model.summary())
model.fit_generator(train, epochs=10, validation_data=validation, verbose=1)
loss, accuracy = model.evaluate_generator(test, verbose=0)
print(f'Model Accuracy: {accuracy}')

Turns out this specific issue relates to a patch that Quantconnect made to pandas dataframes which interfered with the older version of tensorflow/keras.

Tensorflow Keras MNIST get probabilities of prediction [duplicate]

This question already has answers here:
Returning probabilities in a classification prediction in Keras?
(2 answers)
Closed 1 year ago.
I have trained my model and it works fine. Then I went on to predict single images (jpg). This also works, but I don't get the exact probabilities now.
This is my model:
def train():
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
#nomalize data
x_train = tf.keras.utils.normalize(x_train, axis=1)
x_test = tf.keras.utils.normalize(x_test, axis=1)
#train model
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(128, activation=tf.nn.relu))
model.add(tf.keras.layers.Dense(128, activation=tf.nn.relu))
model.add(tf.keras.layers.Dense(10, activation=tf.nn.softmax))
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=3)
model.save('Mnist')
print("Done with training :)")
And this is how I predict my single image:
def predict(testImg):
import numpy as np
model = load_model('Mnist')
img = testImg.convert('L').resize((28,28), Image.ANTIALIAS)
img = np.array(img)
predictions = model.predict(img[None,:,:])
I am suspecting it has something to do with the img[None,:,:] reshaping because the predict function was giving back the probabilities with my test set.
Now I am just getting back an array like [0,0,0,0,0,1,0,0,0] and not actual probabilities.

It turned out I was missing these lines before I made my prediction to normalize my array.
img = img.astype('float32')
img /= 255
As also discribed here:
Returning probabilities in a classification prediction in Keras?

See all correctly and incorrectly identified images when training on the mnist dataset

I'm trying to find a way to visualize which numbers in the mnist dataset a model was able to correctly identify and which ones it wasn't.
What I can't seem to find is if such a visualization is possible in tensorboard or if I would need to use/create something else to achieve it.
I'm currently working from the basic tutorial provided for tensorflow 2.0 with tensorboard added.
import datetime
import tensorflow as tf
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10, activation='softmax')
])
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
log_dir="logs/fit/" + datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)
model.fit(x_train,
y_train,
epochs=5,
validation_data=(x_test, y_test),
callbacks=[tensorboard_callback])
model.evaluate(x_test, y_test)

It appears the what-if tool is what I was looking for, it allows you to visually sort testing data depending on whether it was correctly or incorrectly identified by the model.
If you want to test it out here is their demo that I used to get the above image and they have multiple other demos on the tools site.

Keras ValueError when loading weights

This is the error message I got
Traceback (most recent call last):
File "/home/xxx/Documents/program/test.py", line 27, in <module>
model.load_weights('models/model.h5')
File "/home/xxx/Documents/program/venv/lib/python3.6/site-packages/tensorflow/python/keras/engine/network.py", line 1391, in load_weights
saving.load_weights_from_hdf5_group(f, self.layers)
File "/home/xxx/Documents/program/venv/lib/python3.6/site-packages/tensorflow/python/keras/engine/saving.py", line 732, in load_weights_from_hdf5_group
' layers.')
ValueError: You are trying to load a weight file containing 2 layers into a model with 0 layers.
From this minimal example that produces the error
from tensorflow import keras
from data import get_data
X_train, y_train, X_val, y_val = get_data() # get some train and val data
model = keras.Sequential()
model.add(keras.layers.Dense(64, activation='relu'))
model.add(keras.layers.Dense(7, activation='softmax'))
model.compile(
optimizer=keras.optimizers.Adam(1e-4),
loss='categorical_crossentropy',
metrics=['accuracy']
)
model.fit(
x=X_train,
y=y_train,
batch_size=500,
epochs=200,
verbose=2,
validation_data=(X_val, y_val)
)
model.save_weights('models/model.h5')
model.load_weights('models/model.h5')
Directly running this does not produce the error. However, when I run the program for a second time commenting out the training part (from line 10 to 25) trying to load the weights, it gives me this error.
I am using Tensorflow 1.9.0 and the built-in Keras.

As mentioned above, there seems to be a bug in keras sequential mode: https://github.com/keras-team/keras/issues/10417.
However, you can get around this by using the Keras Functional API (you'll also find the Functional API much more useful when you're building trickier RNNs models with complicated I/O and tensor concatenations).
The disadvantage of using model.save_weights() method to save your neural network is that you have to invoke the model architecture before you load .h5 weights into the NN. If you instead save the whole model (both parameters AND architecture), you'll find that it's much easier to load trained model into a Python object. You can achieve this by using model.save() method.
### TRAINING CODE
import tensorflow as tf
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
# some data
iris = load_iris()
X, y = iris.data, iris.target
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2)
y_train_oh = tf.keras.utils.to_categorical(y_train)
y_val_oh = tf.keras.utils.to_categorical(y_val)
# Keras Functional API
x = tf.keras.Input(shape=(4,))
dense = tf.keras.layers.Dense(64, activation='relu')(x)
dense = tf.keras.layers.Dense(3, activation='softmax')(dense)
model = tf.keras.Model(inputs=x, outputs=dense)
model.compile(optimizer=tf.keras.optimizers.Adam(1e-4),
loss='categorical_crossentropy',
metrics=['accuracy'])
# training
model.fit(X_train, y_train_oh, 16, epochs=20, validation_data=(X_val, y_val_oh))
# save weights
model.save_weights('models/model_weights.h5')
# save weights AND architecture
model.save('models/model.h5')
### TESTING CODE
# Model loading using .h5 weights file
import tensorflow as tf
x = tf.keras.Input(shape=(4,))
dense = tf.keras.layers.Dense(64, activation='relu')(x)
dense = tf.keras.layers.Dense(3, activation='softmax')(dense)
model2 = tf.keras.Model(inputs=x, outputs=dense)
model2.load_weights('models/model_weights.h5')
# Model loading using .h5 model file
import tensorflow as tf
model3 = tf.keras.models.load_model('models/model.h5') # simpler API, but bigger filesize

Loading a trained Keras model and continue training

I was wondering if it was possible to save a partly trained Keras model and continue the training after loading the model again.
The reason for this is that I will have more training data in the future and I do not want to retrain the whole model again.
The functions which I am using are:
#Partly train model
model.fit(first_training, first_classes, batch_size=32, nb_epoch=20)
#Save partly trained model
model.save('partly_trained.h5')
#Load partly trained model
from keras.models import load_model
model = load_model('partly_trained.h5')
#Continue training
model.fit(second_training, second_classes, batch_size=32, nb_epoch=20)
Edit 1: added fully working example
With the first dataset after 10 epochs the loss of the last epoch will be 0.0748 and the accuracy 0.9863.
After saving, deleting and reloading the model the loss and accuracy of the model trained on the second dataset will be 0.1711 and 0.9504 respectively.
Is this caused by the new training data or by a completely re-trained model?
"""
Model by: http://machinelearningmastery.com/
"""
# load (downloaded if needed) the MNIST dataset
import numpy
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import np_utils
from keras.models import load_model
numpy.random.seed(7)
def baseline_model():
model = Sequential()
model.add(Dense(num_pixels, input_dim=num_pixels, init='normal', activation='relu'))
model.add(Dense(num_classes, init='normal', activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model
if __name__ == '__main__':
# load data
(X_train, y_train), (X_test, y_test) = mnist.load_data()
# flatten 28*28 images to a 784 vector for each image
num_pixels = X_train.shape[1] * X_train.shape[2]
X_train = X_train.reshape(X_train.shape[0], num_pixels).astype('float32')
X_test = X_test.reshape(X_test.shape[0], num_pixels).astype('float32')
# normalize inputs from 0-255 to 0-1
X_train = X_train / 255
X_test = X_test / 255
# one hot encode outputs
y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)
num_classes = y_test.shape[1]
# build the model
model = baseline_model()
#Partly train model
dataset1_x = X_train[:3000]
dataset1_y = y_train[:3000]
model.fit(dataset1_x, dataset1_y, nb_epoch=10, batch_size=200, verbose=2)
# Final evaluation of the model
scores = model.evaluate(X_test, y_test, verbose=0)
print("Baseline Error: %.2f%%" % (100-scores[1]*100))
#Save partly trained model
model.save('partly_trained.h5')
del model
#Reload model
model = load_model('partly_trained.h5')
#Continue training
dataset2_x = X_train[3000:]
dataset2_y = y_train[3000:]
model.fit(dataset2_x, dataset2_y, nb_epoch=10, batch_size=200, verbose=2)
scores = model.evaluate(X_test, y_test, verbose=0)
print("Baseline Error: %.2f%%" % (100-scores[1]*100))
Edit 2: tensorflow.keras remarks
For tensorflow.keras change the parameter nb_epochs to epochs in the model fit. The imports and basemodel function are:
import numpy
from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.models import load_model
numpy.random.seed(7)
def baseline_model():
model = Sequential()
model.add(Dense(num_pixels, input_dim=num_pixels, activation='relu'))
model.add(Dense(num_classes, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model

Actually - model.save saves all information need for restarting training in your case. The only thing which could be spoiled by reloading model is your optimizer state. To check that - try to save and reload model and train it on training data.

Most of the above answers covered important points. If you are using recent Tensorflow (TF2.1 or above), Then the following example will help you. The model part of the code is from Tensorflow website.
import tensorflow as tf
from tensorflow import keras
mnist = tf.keras.datasets.mnist
(x_train, y_train),(x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
def create_model():
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(512, activation=tf.nn.relu),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10, activation=tf.nn.softmax)
])
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy',metrics=['accuracy'])
return model
# Create a basic model instance
model=create_model()
model.fit(x_train, y_train, epochs = 10, validation_data = (x_test,y_test),verbose=1)
Please save the model in *.tf format. From my experience, if you have any custom_loss defined, *.h5 format will not save optimizer status and hence will not serve your purpose if you want to retrain the model from where we left.
# saving the model in tensorflow format
model.save('./MyModel_tf',save_format='tf')
# loading the saved model
loaded_model = tf.keras.models.load_model('./MyModel_tf')
# retraining the model
loaded_model.fit(x_train, y_train, epochs = 10, validation_data = (x_test,y_test),verbose=1)
This approach will restart the training where we left before saving the model. As mentioned by others, if you want to save weights of best model or you want to save weights of model every epoch you need to use keras callbacks function (ModelCheckpoint) with options such as save_weights_only=True, save_freq='epoch', and save_best_only.
For more details, please check here and another example here.

The problem might be that you use a different optimizer - or different arguments to your optimizer. I just had the same issue with a custom pretrained model, using
reduce_lr = ReduceLROnPlateau(monitor='loss', factor=lr_reduction_factor,
patience=patience, min_lr=min_lr, verbose=1)
for the pretrained model, whereby the original learning rate starts at 0.0003 and during pre-training it is reduced to the min_learning rate, which is 0.000003
I just copied that line over to the script which uses the pre-trained model and got really bad accuracies. Until I noticed that the last learning rate of the pretrained model was the min learning rate, i.e. 0.000003. And if I start with that learning rate, I get exactly the same accuracies to start with as the output of the pretrained model - which makes sense, as starting with a learning rate that is 100 times bigger than the last learning rate used in the pretrained model will result in a huge overshoot of GD and hence in heavily decreased accuracies.

Notice that Keras sometimes has issues with loaded models, as in here.
This might explain cases in which you don't start from the same trained accuracy.

You might also be hitting Concept Drift, see Should you retrain a model when new observations are available. There's also the concept of catastrophic forgetting which a bunch of academic papers discuss. Here's one with MNIST Empirical investigation of catastrophic forgetting

All above helps, you must resume from same learning rate() as the LR when the model and weights were saved. Set it directly on the optimizer.
Note that improvement from there is not guaranteed, because the model may have reached the local minimum, which may be global. There is no point to resume a model in order to search for another local minimum, unless you intent to increase the learning rate in a controlled fashion and nudge the model into a possibly better minimum not far away.

If you are using TF2, use the new saved_model method(format pb). More information available here and here.
model.fit(x=X_train, y=y_train, epochs=10,callbacks=[model_callback])#your first training
tf.saved_model.save(model, save_to_dir_path) #save the model
del model #to delete the model
model = tf.keras.models.load_model(save_to_dir_path)
model.fit(x=X_train, y=y_train, epochs=10,callbacks=[model_callback])#your second training

It is completely okay to train a model with a saved model. I trained the saved model with the same data and found it was giving good accuracy. Moreover, the time taken was quite less in each epoch.
Here is the code have a look:
from keras.models import load_model
model = load_model('/content/drive/MyDrive/CustomResNet/saved_models/model_1.h5')
history=model.fit(train_gen,validation_data=valid_gen,epochs=5)