PyTorch - Transformer code walkthrough - Part 2 Actual Code

preview_player
Показать описание
Transformer code walkthrough:
Рекомендации по теме
Комментарии
Автор

thank you for the video. is there any github for the code?

somayehseifi
Автор

you have the elements to implement it with the video, but content is very disorganized. Need to watch several times to understand.

alfredoejaramillo
Автор

Thanks for this code along, very useful. But I think the model doesn't make any sense. The guy who made this code basically stacked a lot of unrelated sentences together by first concatenating every single token and then batchfying. In this way the model will end up learning to orient the attention mechanism over a certain word span, by including words that are absolutely not related to each other because originally they were not in the same sentence.

marcod
welcome to shbcf.ru