Convolutional Neural Networks in TensorFlow week 1 Assigment

Course Q&A TensorFlow Developer Professional Convolutional Neural Networks in TensorFlow
1

When creating a post, please add:

  • Link to the classroom item you are referring to:
    I don’t know what I am doing wrong with the week 1 assigment. When the model is trained is taking so long that never ends. It is taking more than 30 minutes to execute and then the google collab runs out of time. I am using at least 3 convolution layers and everything is working out perfect until I need to train the model.
2

It is part of the Tensorflow Professional Certification. Course 2: Convolutional Neural Netwirks in TensorFlow, week 1. The Assignment is creating the convolutional Neural Networks, a cat and dog classifier. When using the training data: 12500 images for dogs and 12500 images for cats, it never ends.

3

Please confirm the following:

  1. Using a GPU for training the model.
  2. No augmentations are performed on the train dataset.

Training time for this assignment is influenced by 2 factors:

  1. Model architecture: Look for the number of parameters the model is training on using model.summary(). One way to speed up training is to reduce the number of parameters (by using more conv* and *pool layers) since the learnable parameters on a Dense layer goes down.
  2. (Probably out of your hands in this assignment but is good to know) Stop relying on ImageDataGenerator (read this)