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

Add orthogonal initialization #1586

Open
noahfarr opened this issue Nov 13, 2024 · 4 comments · May be fixed by #1651
Open

Add orthogonal initialization #1586

noahfarr opened this issue Nov 13, 2024 · 4 comments · May be fixed by #1651
Labels
enhancement New feature or request good first issue Good for newcomers

Comments

@noahfarr
Copy link
Contributor

noahfarr commented Nov 13, 2024

Hi,

I would like to use orthogonal initialization as I would like to use it for a project of mine. Is it planned to add it in the near future?

@awni
Copy link
Member

awni commented Nov 13, 2024

There is no plan that I know of to add it. If you are up for it you can send a PR. Otherwise we can add it soon.. I don't think it should be too difficult.

@awni awni added enhancement New feature or request good first issue Good for newcomers labels Nov 13, 2024
@noahfarr
Copy link
Contributor Author

Its not trivial as you need e.g. qr decomposition and idk if thats implemented in mlx, so you either have to use some third party package or implement that as well

@awni
Copy link
Member

awni commented Nov 13, 2024

We have a QR decomposition. The main limitation is it runs on the CPU.. so that part of the initialization needs to run on the CPU.

@noahfarr
Copy link
Contributor Author

Ah very nice. Then I maybe will put together a pull request in a few days

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request good first issue Good for newcomers
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants