All Questions
Tagged with tf.keras tensorflow-estimator
17
questions
11
votes
3
answers
37k
views
Tensorflow==2.0.0a0 - AttributeError: module 'tensorflow' has no attribute 'global_variables_initializer'
I'm using Tensorflow==2.0.0a0 and want to run the following script:
import tensorflow as tf
import tensorboard
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import ...
- 4,916
4
votes
1
answer
985
views
Tensorflow model analysis, TFMA for keras model
I would like to use TFMA with keras model. The keras model was created with TF 2.0 alpha.
The model is a pretrained model with a classification layer:
...
- 679
4
votes
0
answers
250
views
Tensorflow/Keras local training 6x faster than distributed training
I am training a Classifier (via RNN) in Tensorflow with tf.data api, tf.keras api and tf. estimator api. In local training i got a troughput of around 240 steps/sec, than i enter distributed mode with ...
- 69
3
votes
1
answer
410
views
TF 2.0 - Method estimator.model_to_estimator( ) fails but model.fit works for tf.keras created model
While using the TF 2.0 Alpha and converting a working TF Keras model to TF estimator I encountered a strange error.
The below code has been taken from the Tensorflow without PHD series - RNN time ...
- 81
2
votes
2
answers
414
views
Project organization with Tensorflow.keras. Should one subclass tf.keras.Model?
I'm using Tensorflow 1.14 and the tf.keras API to build a number (>10) of differnet neural networks. (I'm also interested in the answers to this question using Tensorflow 2). I'm wondering how I ...
- 1,473
2
votes
2
answers
931
views
How to build input function for a estimator created from keras model
I'm creating an estimator from keras model as below
estimator = tf.keras.estimator.model_to_estimator(keras_model=keras_model, model_dir=model_dir)
my model is like
...
- 114
2
votes
0
answers
600
views
"Epochs" instead of "Steps" in the horizontal axis in tensorboard when using model_to_estimator?
I am using keras_model_to_estimator to train an estimator and use tensorboard for analysis. The estimator logs the steps and not epochs, and hence the same appears in the x-axis in tensorboard.
I ...
2
votes
0
answers
591
views
How to set loss for train_op using tf estimator API when the loss is input-dependent and implemented via a custom loss layer?
I am implementing a CNN model with tf.estimator API.
Briefly, the cost/loss function is input-dependent and contains variables to be updated during training. I implemented a custom layer with ...
- 21
1
vote
2
answers
155
views
Tensorflow Keras Estimator fails on regression task while underlying model works
I use a convolutional neural network for a regression task (i.e. the final layer of the network has one neuron with linear activation) and this works fine (enough). When I try to use the exact same ...
- 1,473
1
vote
1
answer
78
views
Dimensions Mismatch during Clone Model step of Final Export of Estimator
I'm creating a Tensorflow Estimator from a Keras model. Currently, the estimator is created, the model is trained, and the model is evaluated without issues. However, on the last evaluation, the model ...
- 167
1
vote
0
answers
269
views
tf.keras.estimator.model_to_estimator failing to convert keras model with custom and Lambda layers
I've written a model some time ago that uses a few custom layer definitions and has been trained using TF 1.12 and standalone Keras 2.2.4. I've updated the version of TF to 1.14 and switched over to ...
- 229
1
vote
1
answer
793
views
Building models and estimators in tf2.0 without tf.keras
Given that layers API has been deprecated, how do I build models in tf2 without using tf.keras (or what is the recommended way to build models)? Issue #30829 has the same question, but was closed ...
1
vote
0
answers
90
views
Keras model_to_estimator doesn't log metrics on estimator.train, but logs on estimator.evaluate
I tried the keras model_to_estimator to build an estimator, but found that it wasn't logging metrics during training, although it does during evaluation.
If that doesn't work out of the box, how can ...
1
vote
0
answers
282
views
Tensorflow estimator fails to converge on model converted from Keras (when using binary_crossentropy)
I've been stuck for quite a while using the model_to_estimator functionality in Tensorflow Estimators. The problem seems to be that Keras allows a binary_crossentropy loss on a single neuron Dense ...
- 105
0
votes
1
answer
359
views
Two sets of shared embeddings from one tensorflow feature?
How can I create two sets of shared embeddings from the same tensorflow feature columns?
This small example
import tensorflow as tf
data = {"A": [0, 1, 2, 3], "B": [2, 1, 0, 3]}
...
- 2,017
0
votes
1
answer
469
views
Why does this simple tf.keras model not train after converting to a tensorflow estimator?
I'm trying to convert a tf.keras model to a tensorflow estimator using tf.keras.estimator.model_to_estimator, but the resulting estimator doesn't appear to be trainable.
I've tried modelling y = (x_1 ...
- 11
0
votes
1
answer
1k
views
How to get TensorFlow Estimator version using Python code (not CLI)
I can get the versions of TensorFlow and TensorFlow Keras using these code lines:
import tensorflow as tf;
print("TensorFlow version:",tf.__version__);
print("TensorFlow Keras version:",tf.keras....
- 7,997