# Course5 Week1 assignment 3, only easy if you know the answer

**URL:** <https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134>\
**Category:** Sequence Models\
**Tags:** coursera-platform\
**Created:** [May 25, 2021, 1:25pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134 "2021-05-25T13:25:36Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![biks](https://avatars.discourse-cdn.com/v4/letter/b/6de8d8/32.png) [@biks](https://community.deeplearning.ai/u/biks)\
**Post date:** [May 25, 2021, 1:25pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/1 "2021-05-25T13:25:36Z")

</div>

Assignment 3 is really hard to understand. It seems to be pure magic and luck when i finally find the right mix of function calls to apply… Maybe a better explanation of how the tensor shapes etc propagate through a Keras functional would make sense? I am really stuck at some weird shape /input problem for “Exercise 2 - music\_inference\_model”

Test failed Expected value [‘InputLayer’, [(None, 64)], 0] does not match the input value: [‘Reshape’, (None, 1, 90), 0]

I do not know if InputLayer is supposed to be Reshape or the other way round…

and after too long googling and code review I am about to stop for today, no solution in sight, super frustrated.

---

<div class="post-metadata">

**Author:** ![TMosh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/tmosh/32/9321_2.png) [@TMosh](https://community.deeplearning.ai/u/TMosh)\
**Post date:** [May 25, 2021, 3:56pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/2 "2021-05-25T15:56:21Z")

</div>

Try this thread:  
[Jazz tips](https://community.deeplearning.ai/t/jazz-improvisation-with-lstm-problem/3901/38)

---

<div class="post-metadata">

**Author:** ![akhtarh](https://avatars.discourse-cdn.com/v4/letter/a/9f8e36/32.png) [@akhtarh](https://community.deeplearning.ai/u/akhtarh)\
**Post date:** [May 29, 2021, 4:46pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/3 "2021-05-29T16:46:55Z")

</div>

I am facing the same problem, even given thread doesn’t help much.

---

<div class="post-metadata">

**Author:** ![dbiber](https://avatars.discourse-cdn.com/v4/letter/d/c5a1d2/32.png) [@dbiber](https://community.deeplearning.ai/u/dbiber)\
**Post date:** [May 29, 2021, 5:10pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/4 "2021-05-29T17:10:57Z")

</div>

I started a thread one this last night [Course 5 Week 1 Project 3 - #7 by dbiber](https://community.deeplearning.ai/t/course-5-week-1-project-3/10044/8).

Got a bit of feedback from a mentor on a simple oversight I made. Then used it to kind of track my errors i was running into. I get confused when I am writing after being zoned into this for +12 hours lol so sorry if my troubleshooting seems to make no sense, but it the end i am very close to having this thing locked down. Also by aware since they just released the updated platform there are bugs. I have quite a few with the notebook saying everything is good and passes. But the overall autograder with the hidden tests fails.

---

<div class="post-metadata">

**Author:** ![TMosh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/tmosh/32/9321_2.png) [@TMosh](https://community.deeplearning.ai/u/TMosh)\
**Post date:** [May 29, 2021, 8:00pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/5 "2021-05-29T20:00:24Z")

</div>

Debugging tip:  
If you can’t fix a problem in under an hour, best practice is to walk away and do something else, then come back later with a fresh mind.

Twelve hours of doing the same thing is just going to lead to confusion and chasing a lot of false leads.

---

<div class="post-metadata">

**Author:** ![TMosh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/tmosh/32/9321_2.png) [@TMosh](https://community.deeplearning.ai/u/TMosh)\
**Post date:** [May 29, 2021, 8:06pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/6 "2021-05-29T20:06:07Z")

</div>

There are three classes of errors regarding failure to score 100 when you submit.

- The built-in test cases do not check your code for every possible error. You can easily pass the test cases with defective code.
- The Notebook platform isn’t all that reliable - it can sometimes get confused and not include your latest changes to your code cells. That’s why we recommend you restart the kernel and re-run all of the cells.
- Sometimes your code will cause the grader to actually break. The grader isn’t perfectly immune to coding mistakes.

It’s very difficult for anyone (students or mentors) to identify which of those three situations exist in a given situation.

---

<div class="post-metadata">

**Author:** ![dbiber](https://avatars.discourse-cdn.com/v4/letter/d/c5a1d2/32.png) [@dbiber](https://community.deeplearning.ai/u/dbiber)\
**Post date:** [May 29, 2021, 9:39pm UTC](https://community.deeplearning.ai/t/course5-week1-assignment-3-only-easy-if-you-know-the-answer/9134/7 "2021-05-29T21:39:34Z")

</div>

You are absolutely right. I got pretty brunt out. So I am just going to be working out in sparingly. I’ve actually spent about 12 hours a day since the middle of march to build me resume up till I found employment. I found it though so just finishing up my open classes before i start.
