facebookresearch / vissl

VISSL is FAIR's library of extensible, modular and scalable components for SOTA Self-Supervised Learning with images.
https://vissl.ai
MIT License
3.26k stars 334 forks source link

Possibility of adding weights *AFTER* Imagenet fine-tuning? #430

Open as3895 opened 3 years ago

as3895 commented 3 years ago

🚀 Feature

Sharing weights after the networks have been fine-tuned on ImageNet.

Motivation & Examples

Currently, the repo lists the models and reports their corresponding ImageNet top-1 accuracy. Though, they do provide the scripts to fine-tune a specific model on ImageNet it would be immensely helpful to the community if the weights after the fine-tuning are also available. Fine-tuning on each approach takes considerable time even with excess of GPUs and since these experiments have already been done it might be easy to share these weights.

prigoyal commented 3 years ago

thank you for reaching out. We will look into this request. cc @iseessel if this can be included as part of next vissl release to update the model zoo.