Skip to content
This repository has been archived by the owner on Mar 15, 2024. It is now read-only.

Fine-tuning details on iNaturalist dataset #105

Closed
Reinhard-Tichy opened this issue Jul 30, 2021 · 4 comments
Closed

Fine-tuning details on iNaturalist dataset #105

Reinhard-Tichy opened this issue Jul 30, 2021 · 4 comments
Labels
question Further information is requested

Comments

@Reinhard-Tichy
Copy link

In #45 , I have found some basic details about finetuning for CIFAR10 and Cars, but I could not find anything about other datasets.

So, could you please also provide the details about hyperparameters used in iNat2018 & 2019 for finetuning?

Thank you very much!

@TouvronHugo
Copy link
Contributor

Hi @Reinhard-Tichy ,
Thanks for your question,
The hparams for iNaturalist with the base model are:

  • Batch size: 768
  • LR: 7.5e-5
  • epochs: 360
  • Weight decay 0.05
  • Scheduler: cosine
  • Repeated augmentation: True
  • Smoothing: 0.1
  • Warmup: 5 epochs
  • RandAugment: rand-m9-mstd0.5-inc1
  • Mixup: 0.8
  • CutMix: 1.0
  • Stochastic Depth:0.1
  • Erasing: 0.1
  • Optimizer: AdamW
    Don't hesitate if you have any other questions,
    Best,
    Hugo

@TouvronHugo TouvronHugo added the question Further information is requested label Jul 30, 2021
@Reinhard-Tichy
Copy link
Author

Thanks for your reply! Could you further explain what the param Stochastic Depth means here? Cause I seemly cannot find this arg in the main.py

And does the finetuning process really need a 360 epochs? This is even more than the one of the pretraining step [Lol]

@TouvronHugo
Copy link
Contributor

Hi @Reinhard-Tichy ,
For Stochastic Depth this is the --drop-path argument.
Yes, this allows for better results but it is possible to fine-tune with much less epochs but the results will be slightly worse.
Best,
Hugo

@TouvronHugo
Copy link
Contributor

FYI, I added the SGD config for iNat finetuning here

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

No branches or pull requests

2 participants