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

Added ViT tutorial #102

Merged
merged 1 commit into from
Nov 20, 2024
Merged

Added ViT tutorial #102

merged 1 commit into from
Nov 20, 2024

Conversation

vfdev-5
Copy link
Contributor

@vfdev-5 vfdev-5 commented Nov 19, 2024

@jakevdp jakevdp self-assigned this Nov 19, 2024
Copy link
Collaborator

@jakevdp jakevdp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks great overall! Since the final accuracy is so low, I wonder if it would be worth augmenting this tutorial with some fine-tuning to improve the model? That's something we've alluded to so far in the examples, but haven't yet really dug into

docs/JAX_Vision_transformer.md Outdated Show resolved Hide resolved
@vfdev-5
Copy link
Contributor Author

vfdev-5 commented Nov 19, 2024

I wonder if it would be worth augmenting this tutorial with some fine-tuning to improve the model?

In this case we should fetch trained weights from somewhere, map and load them.

Maybe, simply I can train for longer and get better results...

@vfdev-5
Copy link
Contributor Author

vfdev-5 commented Nov 20, 2024

@jakevdp I tried various configs, now reducing the model size, image size to 120px and still training for 50 epochs I could get 67% accuracy. Maybe, it is OK for now.

Copy link
Collaborator

@jakevdp jakevdp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

@jakevdp jakevdp merged commit fd49eae into jax-ml:main Nov 20, 2024
6 checks passed
@vfdev-5 vfdev-5 deleted the vit-tutorial branch November 25, 2024 12:19
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

Successfully merging this pull request may close these issues.

2 participants