All Questions
1,097 questions
0
votes
1
answer
63
views
Why do model.evaluate() vs. manual loss computation with model.predict() in tf.keras do not add up?
I use keras and tensorflow to train a 'simple' Multilayer Perceptron (MLP) for a regression task, where I use the mean-squared error (MSE) as loss-function. I denote my training data as x_train, ...
- 113
1
vote
1
answer
91
views
numpy eager execution problem after loading a CNN model
I want to save and load a CNN model to further training. I have developed a model and saved it as .h5 file. There is no problem when creating, training, and saving at first run.
The problem exists ...
0
votes
1
answer
113
views
Training a Keras model to identify leap years
I'm trying to teach my learning model how to identify leap years.
My Python code can be seen below:
import tensorflow as tf
import numpy as np
import calendar
# ------------------- UTILITY FUNCTIONS -...
- 11
4
votes
1
answer
46
views
keras model does not learn if using tf.data pipeline
A keras model does learn using Numpy arrays as the input, but fails to make any progress if reading data from a tf.data pipeline. What could be the reason potentially?
In particular, the model ...
- 2,552
0
votes
1
answer
43
views
How to apply a threshold filter to a layer?
Having an array like this :
input = np.array([[0.04, -0.8, -1.2, 1.3, 0.85, 0.09, -0.08, 0.2]])
I want to change all the values (of the last dimension) between -0.1 and 0.1 to zero and change the ...
- 4,864
0
votes
0
answers
111
views
I need to convert a symbolic Keras tensor into a NumPy array to pass it via PyBind11 to a C++ lib
I use Python 3.12.3, NumPy 1.26.4, TensorFlow 2.17.0 - this one without GPU support.
According to some research, these are compatible. I've installed them in a private environment in my home directory....
- 29
0
votes
1
answer
53
views
What is the difference between Dropout(1.0) and stop_gradient?
Consider these two architectures:
prev_layer -> dropout 1.0 -> next_layer (output layer)
prev_layer -> stop_gradient -> next_layer (output layer)
As gradients flow from the output layer ...
- 343
0
votes
1
answer
46
views
How do I pass sklearns train_test_split actual dataseries and not single values as input argument?
I want to train an LSTM-based RNN model for binary classification and for that I wanted to use tensorflow keras model with LSTM layers. In order to do so, I need testing input and output as well as ...
- 105
0
votes
1
answer
53
views
TensorFlow - InvalidArgumentError: Shapes of all inputs must match
I am currently following the TensorFlow guide/tutorial on seq2seq NMT models (https://www.tensorflow.org/text/tutorials/nmt_with_attention) using a Jupyter Notebook.
Upon running the following code,
# ...
- 1,338
0
votes
1
answer
42
views
Preventing weights update for certain samples in Keras3
I have a custom layer called ControlledBackward that accepts prev_layer and input_layer_of_mask:
class ControlledBackward(Layer):
def __init__(self, **kwargs):
super(ControlledBackward, self)....
- 343
3
votes
1
answer
1k
views
Hashed cross-product transformation in PyTorch
I want to implement a hashed cross product transformation like the one Keras uses:
>>> layer = keras.layers.HashedCrossing(num_bins=5, output_mode='one_hot')
>>> feat1 = np.array([1, ...
- 706
0
votes
1
answer
41
views
arr_to_image changes the grayscale images
When processing images, I encountered strange behavior when converting a PIL gray image to a NumPy array.
Here is a small code that loads a gray image image.
from keras.utils import img_to_array, ...
- 1
0
votes
0
answers
29
views
Reshaping spatiotemporal dimension of mmap numpy so that it can be fed to ConvLSTM2D
I have this tensor of multidimensional array:
# read from memory-mapped files
# File paths for memory-mapped files
input_file_path = 'tensor_input.dat'
output_file_path = 'tensor_output.dat'
# ...
- 1,162
0
votes
0
answers
756
views
Getting Error "object __array__ method not producing an array" while training a model
I am trying to train a model using the neural network model Sequential. Every thing is going good till the model compilation. As soon as i hit model.fit() for model training i get the error "...
1
vote
1
answer
41
views
Broadcasting multiple versions of X_data that pair with the same y_data
My deep learning architecture accepts an input vector with size 512 and an output vector with size 512 too.
The problem is I have X_data version that pairs with the same y_data.
I have these tensors:
(...
- 1,162