Skip to content
This repository has been archived by the owner on Jul 24, 2024. It is now read-only.

Dataset Split for Linear Evaluation and Fine Tuning #184

Open
vamsi231297 opened this issue Jan 11, 2022 · 1 comment
Open

Dataset Split for Linear Evaluation and Fine Tuning #184

vamsi231297 opened this issue Jan 11, 2022 · 1 comment

Comments

@vamsi231297
Copy link

Can someone tell me which of the following methods are being used during the evaluation stage :

  • Doing pretraining on datasplit1 and during the evaluation we use datasplit1 as train and other split as test
  • Doing pretraining on datasplit1 and during the evaluation we use datasplit2 as train and other split as test
@chentingpc
Copy link
Contributor

Both are valid, depends on specific setting. For standard setup, we pretrain on ImageNet and fine-tune / linear eval on ImageNet (eval on their validation set). For transfer learning setup, we pretrain on ImageNet and fine-tune on another dataset (e.g. CIFAR-10 train, and eval on CIFAR-10 test).

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

No branches or pull requests

2 participants