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

Keras Tuner Support #129

Open
saioru opened this issue Jan 11, 2024 · 3 comments
Open

Keras Tuner Support #129

saioru opened this issue Jan 11, 2024 · 3 comments

Comments

@saioru
Copy link

saioru commented Jan 11, 2024

Quite new to Keras trainings, would like to ask if its applicable to leverage keras-tuner on optimizing the hyperparameters in finetune and/or training of the models as this library is a Keras implementation of deepinsight/insightface.

@leondgarse
Copy link
Owner

Ya, it's possible applying keras-tuner here, but needs some works. Like modifying train.py for training hyper parameters, and model architectures in backbones, adding hp configurations. But never have I tried that, not sure the affect on results...
You may also refer All from scratch #71 for a basic usage with keras and tensorflow alone.

@saioru
Copy link
Author

saioru commented Jan 15, 2024

Ya, it's possible applying keras-tuner here, but needs some works. Like modifying train.py for training hyper parameters, and model architectures in backbones, adding hp configurations. But never have I tried that, not sure the affect on results... You may also refer All from scratch #71 for a basic usage with keras and tensorflow alone.

Cheers for the heads up on this, will be conducting a few experiments and modifications to train.py and hp configs.
Any suggested amount of data prep required for fine-tuning ResNet34 with Arcface?

@leondgarse
Copy link
Owner

You may start with 100 classes and 50 images per class. Just a reasonable start point. :)

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

2 participants