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

Training without pretrained weights #6

Open
chinmayembedded opened this issue May 6, 2021 · 1 comment
Open

Training without pretrained weights #6

chinmayembedded opened this issue May 6, 2021 · 1 comment

Comments

@chinmayembedded
Copy link

chinmayembedded commented May 6, 2021

Hi,

I trained training aCASTLE by 1 shot 5 way strategy without using pretrained weights. I trained model for 80 epochs. Performance numbers observed were drastically low compared to training with pretrained weights.

If I want to train a model without pretrained weights, is it possible? or Can you please share methodology to create pretrained weights files.

Thank you!

@Han-Jia
Copy link
Collaborator

Han-Jia commented May 10, 2021

Hi, we pre-train the model with the same strategy as in the FEAT repo. So you may try the pre-trained weights provided by that repo.

Based on our observation, we find the model with pre-trained weights gets much higher GFSL performance than the one without pre-trained weights, but they have similar FSL performance.

BTW, if training from scratch, it may require much longer epochs.

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