Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

The colab notebook is broken again #222

Open
Blazeolmo opened this issue Apr 28, 2023 · 3 comments
Open

The colab notebook is broken again #222

Blazeolmo opened this issue Apr 28, 2023 · 3 comments

Comments

@Blazeolmo
Copy link

Errors when running the first cell. i am impressed, really, that this time it broke so hard not even the first cell works.
I wonder what this project could be if it wasn't made by just a random dude...

@atari486
Copy link

atari486 commented May 7, 2023

Yeah, no one (least of all the developer) has any vested interest in helping. The courtesy of backwards compatibility is long gone, so you can't expect it. Nobody cares anymore. All people care about is being able to do the next new shiny thing, even if it breaks everything that already exists.

With that being a reality, I recommend trying to get in with the free machines on Paperspace instead. They haven't updated all their Pythons/packages (yet) so it'll work if you make some slight modifications (pay attention to the error messages). Of course, advertising this will probably make it even harder to snag one of the free instances now, but better some chance than none.

@tommysteryy
Copy link

If you're looking for, essentially the same notebook, but one that works, I recommend the one linked here instead: https://minimaxir.com/2019/09/howto-gpt2/. This one works totally fine for me on Google Collab and it practically is the same notebook just not using aitextgen. Of course, you lose the other functionality in aitextgen, but it's a good start.

@atari486
Copy link

Thank you Tommy. This is just what I needed, and I can confirm that it works.

Yet I still see this:

  • In the glory days of textgenrnn, we could set anything - dropout rate, RNN size/layers, etc.
  • With the switch to aitextgen, we lost many of those options.
  • Now, you can't even continue training without restarting the runtime.

This is not progress. As time goes on, the users should have more control, not less.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants