PyTorch Datasets: Loading Images and Text for Training

preview_player
Показать описание

In this video we will explore the most important part of training Neural Networks: Passing data from your hard drive to the model! We will be walking through the implementation of a PyTorch dataset for both Computer Vision and Natural Language. Although simplified, these principles will come in handy when implementing much more complex architectures!

Timestamps:
00:00:00 Introduction
00:01:00 Writing a Dataset for Image Classification
00:16:34 Ensuring Images are Stackable for Batches
00:28:40 Train/Test Split Dataset
00:34:14 PyTorch ImageFolder
00:35:56 Datasets for Language Classification
00:36:14 Tokenizing Text (Naive Character-Level)
00:54:27 Writing the Text Dataset Class
01:02:50 Dynamically padding text
01:06:20 Custom Data Collator
01:13:08 Increasing Speed of DataLoaders

Socials!
Рекомендации по теме
Комментарии
Автор

man you are a legend, thank you very much for these videos 🙏🙏

itbaansafwan
Автор

Thank you for the great videos. How can we find the datasets?

omidiamir
Автор

Those videos are pretty helpful. Thank you !! I am in my second year in AI department. Can you give me any advice or a roadmap if it is possible ?

crowley
visit shbcf.ru