C3_W1_Lab2_Transfer Learning

Course Q&A TensorFlow: Advanced Techniques Specialization Advanced Computer Vision with TensorFlow
1

Hi!

Just a few notes about this notebook:

  1. Section Parameters: parameter BATCH_SIZE(==32) is never being used. The actual batch_size later on in model.fit is set to 64.
  2. On the same section, the function plot_metrics says:
    def plot_metrics(metric_name, title, ylim=5) → ylim should be 1 for a better visualization (at least for accuracy).
  3. Section “Train the model” states that training the model “… will take around 20 minutes to complete”, but there are 4 epochs and each epoch takes about 15 minutes → more than an hour…on a GPU in Colab. Maybe it should say: each epoch takes about 15-20 minutes.

Thank you for the course!

2

You can experiment on BATCH_SIZE and choose the one which gives you the best performance.
You are free to pass any possible value to the variable ylim according to your need.
I think the model.fit method displays the time required by each epoch.