brok-bucholtz / project-face-generation Goto Github PK

It might help if it's mentioned how to utilize the function.

Many students might end up adding it directly and try to run the function for every batch. Which might slow down their training perhaps? So clarifying, display it after N iterations might do them good. And this might help reviewers as well. Lots of images increase file size and can be slow to upload depending on their internet connection :)

Also, providing them with code that calculates and prints out loss might be good.

It seems you want the students to try to implement this on their own this time - all the training code. Which seems a bit off considering the previous projects didn't have that. If that is the case, then perhaps a bit more detail on what they can do to help themselves more to get the most out of training (in terms of identifying the problems and what specifically to add in that case) would be good for them. Like printing out loss values as training progresses (not just after each epoch for example) or plotting the loss values [an issue I reported for last week's lessons on GANs in this regard] and being able to relate that to how well the model performs etc.

There multiple typos when passing the value reuse=False to many functions. The code reads: reuse=Fales instead.

In the helper.py when obtaining batches, you are already normalizing the images before yielding them.

I think, perhaps, this should be clarified in the notebook. That you are already normalizing the images, and it's not required to do so again.

This is currently based on the GAN MNIST notebook where it was being done in the notebook itself.

When I was working on this project on my laptop everything worked fine, but then I wanted to try things out on FloydHub. In the first code block, I am having an issue:

ImportError                               Traceback (most recent call last)
<ipython-input-5-5c7df70611aa> in <module>()
      9 """
     10 #from tqdm import tqdm
---> 11 import helper
     12 
     13 helper.download_extract('mnist', data_dir)

/output/helper.py in <module>()
      9 import numpy as np
     10 from PIL import Image
---> 11 from tqdm import tqdm
     12 
     13 

ImportError: No module named 'tqdm'

I checked the helper file and tqdm is in there. Plus, I tried to import tqdm, but nothing worked.

Any idea what's going wrong here?

Missing return tuple element at the model_inputs() function docstring

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data) <---missing element
    """