C4_W4_Lab_3 Getting just only red pictures after epoch4

Course Q&A TensorFlow: Advanced Techniques Specialization Generative Deep Learning with TensorFlow
1

I ran the lab but after epoch 4 the pictures shown are just red. At the other epochs before the gan started to improve on creating faces, but after epoch4 I just contiued to get completly red pictures with a loss of 500 and more.

Has anyone had the same issue?

2

I Don’t know if you noticed this statement before your training loop grader cell

The training will also take a long time (more than 30 minutes) and that is to be expected. If you used the mean loss metric suggested above, train the model until that is down to around 320 before submitting.

So you need to get a loss of 320.

if you go little more down in the assignment, you will also find what number of epochs would be ideal.

The right number of epochs depends on the inherent complexity of your dataset. A good rule of thumb is to start with a value that is 3 times the number of columns in your data. If you find that the model is still improving after all epochs complete, try again with a higher value.

Regards
DP

3

I was just on the lab not the assignment for C4W4 and I wondered that I kept on getting just red pixels after having a good start where you could already see “noisy faces”

Just as my computer produced a Bluescreen and I am now out of GPU time, I didn’t see how the gan was doing after producing just red images.

I will restart the lab as soon I get a TPU again. Let’s see if it’s improving

4

As you have already run your model multiple times. I would suggest to run your model next day as you have already made the corrections today. You had selected an epoch of 200 which you eventually changed to 4. Your ultimate goal is to get loss around 320. For which you might have to again run your model multiple times. So better take rest for today. Start fresh tomorrow.

I had to run this model almost 10 times when I was doing this assignment until I got the desired result. Even I had ran out of GPU time once, another two- three times had network issue, so patience is the key for these assignments.

Wish you al the best!!!

Regards
DP

5

Yup I will, thanks, I will stop for today as it is already 20:00, so time for watching the news.