Hello, I have a problem in exercise " Improve MNIST with convolutions". Could you help me please? @balaji.ambresh
Have a look on this method of tensorflow:
this will help you building your model.
Thank you for creating a public topic.
Please specify the input_shape parameter for the 1st layer of the model to get the parameter count.
The reason behind locking a few jupyter notebook cells is to ensure that you don’t modify cells that aren’t under your control in the grader environment. If you’d like to try things out, here’s a reply that explains how to set a cell as editable.
1 Like
Thanks for your advice.
1 Like
Hello Balaji,
Thanks for your advice. The previous problem was solved. However, even though the results of my code are displayed correctly, the scores shows unexpected error. It may be because my submission time was timed out, as I was stuck on the previous problem for too long. Is there anything I can do in order to turn in this assignment? This is the only assignment I have left to submit for this class.
[
bynmmm
](头像签名)
---- Replied Message ----
From | Balaji Ambresh via DeepLearning.AInotifications@dlai.discoursemail.com |
- | - |
Date | 6/19/2023 17:51 |
To | bynmmm@163.com |
Subject | [DeepLearning.AI] [Introduction to TF for Artificial Intelligence …/TF1 Course 1 Week 3] C1_W3_Convolutional Model |
| balaji.ambresh Mentor - MLEP, DLS
June 19 |
- | - |
Thank you for creating a public topic.
Please specify the input_shape parameter for the 1st layer of the model to get the parameter count.
The reason behind locking a few jupyter notebook cells is to ensure that you don’t modify cells that aren’t under your control in the grader environment. If you’d like to try things out, here’s a reply that explains how to set a cell as editable.
Please click my name and message your notebook as an attachment.
Turns out that if you change an uneditable cell with custom content as a new cell, grader doesn’t like it.
Please refresh your workspace and code at places only where indicated in the notebook.
1 Like
[snippet removed by mentor]
Hello community, I have a slightly different problem in the same assignment on Enhancing MNIST Dataset with Convolutions, Course 1 Week 3. I failed to launch my model to go through all 10 epochs with 99.5% accuracy because of the callback function instance (2nd screenshot). But I am not sure it is callback, I have tried many tweaks like checking for indentations but nothing helped me. Please, give any relevant advice, maybe I am blind, I really do not see any way out ;/
Please provide the shapes of training_images and training_labels
1 Like
Thanks you for answering, I went back to the cell when I reshaped and normalized training_images and rewrote them as advised in the official NumPy portal, but a new problem popped up. I do not grasp why the grader brings up some invisible variables like set_model, which I indeed did not declare. Please give a hint, I feel I am beating a dead horse with that.
Please see here keeping the class to subclass in mind.
1 Like
Thank you really much @balaji.ambresh for pointing out the class to subclass thing, I have rewritten the callback function, removed some redundant lines and it has finally worked !! That is what is called experience, thank you a lot
@balaji.ambresh I have stumbled upon another problem, my array of 28x28 images does not reshape - I have tried writing the reshaped array both as a method like training_images.reshape((60000, 28, 28, 1)) and as a function np.reshape(), but neither worked out. Can you please give me a hint of what am I doing wrong? Thanks in advance!
np.reshape does not modify the function argument. Please provide the correct parameters to the function and capture the return value of the function.
UPDATE!
I have tried different ways to input parameters into the function reshape_and_normalize, but unfortunately it still results in the same output. Maybe I am missing the context because when I looked up various solutions, they mentioned providing arguments like the following - images.reshape(images.shapes[0], 28, 28, 1). But as you had said the reshape func does not modify the function argument.
Could you expand on the correct parameters - do you mean the line of code assigned to the images variable ? Thanks in advance.
Please see this