PyTorch Tutorial 10 - Dataset Transforms

New Tutorial series about Deep Learning with PyTorch!
⭐ Check out Tabnine, the FREE AI-powered code completion tool I use to help me code faster: www.tabnine.com/?... *
In this part we learn how we can use dataset transforms together with the built-in Dataset class.
Apply built-in transforms to images, arrays, and tensors. Or write your own custom Transform classes.
- Dataset Transforms
- Use built-in Transforms
- Implement custom Transforms
Part 10: Dataset Transforms
📚 Get my FREE NumPy Handbook:
www.python-engineer.com/numpy...
📓 Notebooks available on Patreon:
/ patrickloeber
⭐ Join Our Discord : / discord
If you enjoyed this video, please subscribe to the channel!
Official website:
pytorch.org/
Part 01:
• PyTorch Tutorial 01 - ...
Logistic Regression from scratch:
• Logistic Regression in...
Code for this tutorial series:
github.com/patrickloeber/pyto...
You can find me here:
Website: www.python-engineer.com
Twitter: / patloeber
GitHub: github.com/patrickloeber
#Python #DeepLearning #Pytorch
----------------------------------------------------------------------------------------------------------
* This is a sponsored link. By clicking on it you will not have any additional costs, instead you will support me and my project. Thank you so much for the support! 🙏

[@AbdulQayyum-kd3gf]

Excellent tutorial. Keep it up bro

[@ilkayand4]

Thanks a lot. I will share your tutorial with my python related friends. You are so cool.

[@patloeber]

That’s great! Thanks for sharing :)

[@williammichael6365]

Thank you. I finally learned how to make my dateset.

[@patloeber]

That’s great!

[@umesh3876]

Exactly what i was searching for thanks a lot.....

[@Frostbyte-Game-Studio]

hey man
loved your tutorial watched the whole series
thanks for the good content
love your accent btw

[@patloeber]

Glad you enjoy it!

[@tarat.techhh]

Amazing Video plz keep uploading. Cheers

[@patloeber]

Thanks! I will :)

[@rayll8579]

Great tutorial! May I ask what is the benefit of using class? Thank you.

[@itzikgutzcha4779]

Using a class in this way works as a function factory with different default values.
To use this as a function, you would have to constantly input the arguments or hard-code them, or other "less clean" ways.

[@stefanobraghettocatoni1464]

i got an error when try dataset[1]: expected np.ndarray (got numpy.float32) and it ocurrs inside ToTensor class. i s it is solved: return torch.from_numpy(np.asarray(inputs)),torch.from_numpy(np.asarray(labels)). Why is this happening? Thank you

[@Joe-oq2zn]

same here

[@wayneqwele8847]

Thanks, very helpful.

[@patloeber]

Glad you like it!

[@alanzhu7538]

Awesome video! I am wondering why we create our own WineDataset class, rather than using the dataset module directly?

[@donfeto7636]

because here we used dataset from pytorch not ours , if you have your dataset you should do it like we did with WineDataset

[@mohammedy.salemalihorbi1210]

Great tutorial, Thanks a lot

[@patloeber]

Glad you like it

[@swadhindas2208]

Can we define __call__ under @staticmethod to avoid self? (6:12)

[@amarug]

Thank you very much for your tutorials! One question: why make the transforms classes with just one call method, why not just a function and pass that?

[@patloeber]

Your way might also be possible, but this is the PyTorch way of doing this. Their API also uses the __call__ method

[@amarug]

@@patloeber I suspected something like that. Thank you very much!
Viele Dank, grüsse aus der Schweiz.. :)

[@patloeber]

@@amarug Sure! Grüße zurück aus Deutschland ;)

[@lilikoimahalo]

Could you show an example of using this for scaling please?

[@lucycai3356]

Thank you so much!

[@patloeber]

You're welcome!

[@alirezamohseni5045]

thanks alot

[@dhanushsuryavamshi4279]

Hello, I have a question. All the transformations we are doing, it can be done using a method as well. Any particular reason why we are creating class?

[@patloeber]

This is how PyTorch recommends it, then you can for example combine these Transform class objects in transforms.Compose. But for single transformations a function is fine, too

[@fahadaslam820]

Thanks for mazing tutorials. Can you please share implementation for Wine classification using a Feedforward network? Because I follow your tutorial on FFN and implemented for the Wine dataset. unfortunately, I am getting bad results (bad learning - loss is not decreasing as expected)

[@patloeber]

you should try to use a scaler (minmax or standard scaler)

[@fahadaslam820]

@@patloeber Hi, thanks for the suggestion. It actually worked. But I do not understand, what type of scaling I should choose for my data? Is there is any scaling guideline?

[@user-ej9dt6wt1x]

Cool!!!

[@user-fk1wo2ys3b]

Thank you

[@patloeber]

Welcome!

[@ifthikhanmohamed1681]

if self.transform what does it mean

[@Nameless_p]

why do you sometimes call it inputs and other times call it features?

[@AbdulQayyum-kd3gf]

Please can you upload the video on multiclass segmentation in PyTorch, if possible how to make multiclass segmentation data loader in PyTorch

[@patloeber]

one of the next videos will contain multiclass segmentation :)

[@AbdulQayyum-kd3gf]

Thank you so much

[@ArunMozhiDevan04]

"__call__" is depreciated use "__new__" instead