Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Can not predict with exported model.

See original GitHub issue

Bug Description

After fitting my autokeras model, I export it, however, I can not predict nor evaluate with this exported Keras model.

from tensorflow.keras.datasets import mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train.shape # (60000, 28, 28)
y_train.shape # (60000,)
y_train[:3] # array([7, 2, 1], dtype=uint8)

x_train = x_train[:1000]
x_test = x_test[:1000]
y_train = y_train[:1000]
y_test = y_test[:1000]

import autokeras as ak

clf = ak.ImageClassifier(max_trials=3, directory = "/tmp/")
clf.fit(x_train, y_train, epochs=3)
predicted_y = clf.predict(x_test)
clf.evaluate(x_test, y_test)

model = clf.export_model()

predicted_y_2 = model.predict(x_test)

# ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1000, 28, 28)

model.evaluate(x_test, y_test)

# ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1000, 28, 28)

Setup Details

Include the details about the versions of:

OS type and version: Ubuntu 18.04
Python: Python 3.6.10
autokeras: 1.0.0
numpy: 1.18.1

Issue Analytics

State:
Created 4 years ago
Comments:12 (8 by maintainers)

Top GitHub Comments

1reaction

zhulingchencommented, Jan 31, 2020

Hi, before you train on x_train and y_train and test on x_test, did you expand a dimension to the last dimension of x_train and x_test and make it have a shape of (None, 28, 28, 1)?

That’s usually how people do on the datasets to train and test the CNNs with the regular TensorFlow/Keras.

0reactions

haifeng-jincommented, Jul 1, 2020

@zhulingchen Would you like to do a user study with us? Have a 30-minute meeting with me to talk about what do you think of AutoKeras. Your thinking is important to us. Thanks.

Top Results From Across the Web

Export Regression Model to Predict New Data - MathWorks

To exclude the training data, click Export Model and select Export Compact Model. This option exports the model with unnecessary data removed. That...

keras - Autokeras exported model performs not as expected

But when I export and save it as a keras model the model performs really bad. It classifies everything wrongly and the predict-function...

Exporting models for prediction - AI Platform - Google Cloud

To serve predictions from AI Platform Prediction, you must export your trained machine learning model as one or more artifacts. This guide describes...

2. Exporting and deploying a model — IPU TensorFlow ...

For non-Keras models, the export method does not accept the batch-size parameter. ... The function is the same predict function that was exported...

Export TensorFlow models in the SavedModel format

The assets directory is optional. It stores supporting files for the prediction service. The variables directory stores the variables saved by ...