C2W3 Graded Function: Last Output of Inception V3 Model Changed

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

In the Programming Assignment on Coursera: Transfer Learning - Horses vs Humans, I was required to load a fixed weights file for inception v3 model.

I downloaded the weights file from https://storage.googleapis.com/mledu-datasets/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5 per instruction. However, the name of last layer in the inception model was “mixed10” instead of “mixed7” and the output size changed too. Would you double check whether the weights file has been updated? Thank you
The graded function blocked me from passing the programming assignment.

2

Hi @chris.favila is there a possibility something like this update to happen…

3

Hi Michael. mixed_10 is indeed the last layer of the Inception model. For grading purposes, the instructions say to use mixed_7 as the last layer to go into the model you will train. Laurence does something similar in the lectures where he doesn’t pick the last layer but something further up instead. Hope this clarifies it.

4

Hi, Chris

Thank you for your clarification. My new submission was able to pass the grading.

It would be great to update the function signature in the notebook for the future learners. Currently, the instruction is misleading.

# GRADED FUNCTION: output_of_last_layer
def output_of_last_layer(pre_trained_model):
  """
  Gets the last layer output of a model

  Args:
    pre_trained_model (tf.keras Model): model to get the last layer output from

  Returns:
    last_output: output of the model's last layer
  """