Should it be "both must be not longer than 512 tokens for training/512 for eval" a typo?

Course Q&A Natural Language Processing NLP with Attention Models
1

image
image1137×357 77.9 KB

In the explanation, is there a typo? Should it be “both must be not longer than 512 tokens for training/512 for eval”

2

Hi @Amazing_Patrick ,

Disclaimer: I have not taken this NLP specialization.

Based on your screenshot, it really seems like there’s a typo in the instructions.

The instructions say max_lenght = 256 tokens for training, and then the code actually sets 512 as max_lenght.

I am pinging @Mubsi who may indicate the right direction of this post.

Thanks!

Juan

3

Hi @Amazing_Patrick & @Juan_Olano,

Thank you for pointing this out. This has been fixed.

Cheers,
Mubsi

4

image
image1121×296 66.3 KB

It seems it is still “256 tokens”.

5

Hi @Amazing_Patrick,

In your current version of the lab, yes, it is still there.

If you want to see the absolute latest version of the lab where this typo has been fixed, you can click on “Help” on the top right corner, and when the panel opens, click “Get latest version”. But please note, doing this will fetch you the latest version of the assignment and you’d have to redo (copy/paste) your solutions in it.

This change was not “forced” on the learners as fixing this typo has no major effect on the assignment. We don’t want them to re-work on the solutions because of trivial typos.

Cheers,
Mubsi

1 Like
6

@Mubsi Thanks you very much!

1 Like