Hi All,
Could you please support to fix below issue I didn’t make anything as no need to write script :
Thanks for reporting the issue.
Could you please revert the notebook to its base version (after backing it up), and check if you still encounter the same problem?
Hi @Alireza_Saei ,
How to revert it to base version?please?
I tried to revert to May 7 as last checkpoint but it’s not work
You can do it from Checkpoints section (File > Checkpoints).
I did but not work ? what can i do ?
the error is not clear to me even !
HI @paulinpaloalto
please support
If you want to follow Alireza’s advice and get a clean copy, here’s a thread that explains how to do that in detail.
The original error message looks pretty clear. Did you find the line that it is pointing to as being incorrectly indented? Indentation is part of the syntax in python. I checked in my copy of that notebook and I see a line like that in the model function, which was a cell that they just gave to you. But it is modifiable, so perhaps you accidentally changed the code there even though there was no need for you to change anything in that cell.
2 Likes
Thank you , I deleted the old copy and get the updated one from Lab Help.
I’m new in Python i was trying to understand the script by using print statement but i can’t figure out the error msg mean, But Thank issue is fixed .. 
1 Like
Dears,
In the previous question we have been applied forward_propagation_with_dropout by adding below 3 line of codes:
D1 = np.random.rand(A1.shape[0],A1.shape[1])
D1 = (D1 < keep_prob).astype(int)
A1= np.multiply(A1,D1)
A1/=keep_prob
My question how can we applied backward_propagation_with_dropout in one single line code as mentioned in the assignment PFB:
It is given that
# Step 1: Apply mask D1 to shut down the same neurons as during the forward propagation
So, instructions for forward_propagation_with_dropout is:
You have to do the same thing. Multiple D1 but with the derivative of A1.
1 Like
Thanks,
I learned a lots since beginning in this course but i found my self not having the ability to make my own model and choose CNN or RNN or how many hidden layers need also how many unit for each layer ..
In addition , i don’t know how to use what i learned in real life issue ex:self car driving ! or simple issues .
can we say in my current position is very early to get solid base in the fundamentals in AI .
can anyone advise on above ?
If you are taking the DLS courses in sequence, we haven’t even gotten to CNNs and RNNs yet, right? You’re right that there is a lot to learn and it takes more experience beyond just the courses to get to the point that you feel fully capable of using Neural Nets to solve problems on your own. The best advice is just to keep learning and finding experiments to run and problems to solve. There are lots of sources of problems at various levels, e.g. Kaggle competitions, that you can use to get more experience. There is a lot of history on the Kaggle site including past competitions where you can see some of the solutions and learn from them.
I hope that you’re not equating self driving cars with simple issues. Start with something much less complex than autonomous vehicles as the next steps in your learning process. But having said that, you’ll see at least an introduction to some of the computer vision issues that are involved there when you take DLS C4 and learn about YOLO in Week 3.
2 Likes
Experimentation and experience.
Salam @paulinpaloalto
Thank you for your explanations and I will follow your advise and will check Kaggle and try to find some examples about LSTM model as I need it to apply some predictions for KPIs values in my Telecom Industry such as Traffic and Users Count for Specific Technology .I would like to predict the Traffic for the upcoming days, I have the data in Microsoft SQL DB and i would like to integrate deep learning with DB in order to do training/test data and build my forecasting model , So later i can show the predicted values in customers reports.
I checked Kaggle site as per your advice i found simple LSTM model, But what i learned is not used in code such as initialize parameter or Forward_prop or compute cost or backprop .maybe I’m in hurry but this is weird how they build model without using which i learned until now .
maybe these thinks will come in the next weeks in this course . PFB
Agreed
Once you switch to building real models everyone uses an ML platform like TensorFlow or PyTorch. That’s what that model does that you are showing in the screenshot. Once you switch to TF, then that takes care of a lot of the things “under the covers”: e.g. they have default weight initializations, so you only have to write code for that if you need an algorithm different than the default. Similarly, they do backpropagation and gradient descent for you: all you need to do is give it the learning rate and specify whether you want to use Adam and so forth.
If you have not yet dipped your toes into those waters, just stay tuned and you’ll see TF for the first time in Week 3 of this course (DLS C2). Then that will be used extensively in ConvNets (C4) and Sequence Models (C5).
Note that Prof Ng has a pedagogical reason for showing us how the algorithms really work in C1 and C2: if he started by showing us TF, then we would lose a lot of intuition about what is going on. He’ll take a very similar approach in C4 W1 and C5 W1: he first shows us how to write the fundamentals of a ConvNet or RNN in python and numpy and then switches to using TF for building “real” models.
1 Like
Appreciated your support to lighting the way for me, I’ agreed about what you said the strategy that Prof Ng put it for the beginners in DL and how they should to think and understand deeply how TF working on behind , So later they will know what’s going on in Library and how can customized or build their own Model.
Also i have been understand that we can doing DL without to use TF or Pytorch by building our own model as Peer Prof Ng show us in the previous courses.
Dear @paulinpaloalto
Please your support on below issue, I read tutors posts which falling in same my issue but no luck . stuck in this issue more than 1 Hour.
That result is quite a bit larger than the expected value in the case that you have not fixed the intentional errors they gave you in back prop. Are you sure you’ve compared your implementation to the formulas given in the instructions? If so, then it’s probably time to look at your code. We can’t do that directly or in a public thread, but please check your DMs for a message from me.