Assignment 2: Train Your First Model


I enrolled quite late as I simply didn’t know about this gem of a learning platform. Thanks a lot for this course Jovian. I am a complete newb in the field of ML, AI and Data Science. I hope to learn a lot in the coming days and try out other courses as well. These are my 1st and 2nd assignment which I completed just in time. I am really excited to share them with the Jovian community. Hoping for some valuable feedback on the way.

Thanks for your feedback . am Available for u also for next time

Hopefully at some time Sir give a chance to do all our assignments if anyone can’t do it on time

2 Likes

Hi everyone,
First of all thanks @aakashns for the time and effort to create this course!

I hope everyone will find this post helpful for everyone.
If you find it helpful do like and share the post.

Any feedback would be greatly appreciated, here is the link

Good . Best of luck for next tasks

Hi,
Look here: ZERO TO GAN's Assignment 2

Hey Hi Everyone!

Got the loss_value as low as 23.6838, have a look at my model below and any feedback is appreciated.

Thanks

Here is my assignment-2. Sorry for being late submission because of the mid-exam.

why using MSE loss function is giving NaN as loss results? I had to use L1 loss function.


checkout my notebook.

Hey, I m writing the exact code for MNIST dataset as was in lecture. But still it is giving me an error like this.


Help me figure this out

Hi, I don’t know if this happened to someone else. I trained my model and like on my second or third attempt, it gave very good results (val_loss=3763.6506), I was very happy.

Then I started to play with the learning rate and epochs, but then it didn’t improve. The final 3 attempts I decided to go back to the same parameters I was using on the Version 9, which was the best so far. But the val_loss started to increase. Weird, so I decided to present the Version 10 as my Assignment delivery.

I restarted the Model, I even downloaded again the MNIST Dataset and nothing.

Here is my notebook:
Insurance linear regression assignment

According to the error your validation loader is not a Tensor, it is a List, I am not sure where the mistake is but do check where you are defining the val_loader and share the link of the notebook

I saw your results have been good, better than mine actually. Could you explain why you use loss = F.l1_loss(out, targets) ?

1 Like

i have submitted the assignment but evaluation is still pending.so it might mean i didnt pass.can anyone point me out the mistakes and give me feedback on what went wrong. that would help me to do better next time

Since it’s a problem with linear relationships (with some non-linearity as well), I tried with different loss functions for linear regression and they got me similar values of loss.
But my model seemed to perform a little bit better with l1_loss than the others when it comes to prediction. :blush:

You have included ‘charges’ column in input_cols. It won’t give correct results.

Thanks for pointing it out! will update the model by adjusting the inputs

1 Like

Can anyone help me with this?

This is my notebook


output
loss: 46633868.0000

Even though the loss figure looks outrageous, the predictions are fair. Can someone explain why am I getting such a loss? I used mean-squared-error loss.