How You can EASILY create Custom Datasets and Loaders!

Pytorch has some of the best tools to load your data and create datasets on the fly. We will cover examples of creating train, test, and validation datasets and data loaders in both PyTorch and PyTorch lightning. Additionally, we will cover what batching is and why you should be doing it.

Intro: 0:00
Overview: 0:37
Why/What Batching?: 1:34
Dataloader and Dataset MVP: 4:09 
Pytorch Lightning Full Example: 14:21
Pytorch Lightning Model with Dataloader: 19:41

[xiangli]

Thanks! Keep up the good work, buddy!

[muhammad_talha]

hey brother, i just come across to your channel its really great, I love it gonna learn alot from your vids, keep posting <3

[FelLoss]

Hi sandro. Thanks for this tutorial. I found it very easy to follow. However, I'm not sure if I correctly understood what your aim was when using the set_fold method. I wish you could give us more examples about this. Thanks!!!

[roljuye]

Hi Sandro, thank you for your helpful tutorial. How can I test the model on a single tensor with star parameters. I basically want to give the model a tensor and get a predicted label in return, that tells me the star type. Could you please make a short code example?

[liirinen]

How do you write a custom dataloader for image data for an autoencoder that uses two images?

[anas.k]

How can we do when the original data is in a json file. Thank you

[oliviercroissant]

there is only one problem. The dataloader works as described only on Linux
ti does not work on Mac OSX for exemple or on Windows

[curiousmind]

What if my data is in hundreds of csv files...?