chingyaoc / dcl Goto Github PK

Thank you for sharing excellent work!
Could you pls share your implementation codes on DeepMind control suite environments for reinforcement learning?
I'm interested in your work but have trouble reproducing your results.
Thank you very much!

Hi @chingyaoc,

Thanks for the great work!

I wonder how to implement the method of (supervised) unbiased CL. It could be similar to the work : Supervised Contrastive Learning or just supervised cross-entropy loss. Is my conjecture right? Could you pls share me the script or give some comments about the implementation?

Thanks a lot!

Hi,
Thanks for your very well organized codes. I'm now trying to reproduce the pre-training on CIFAR-10, while it seems that I might need to tune the hyper-parameters existed in your STL codes. Could you let me know what will be the parameters that needs to be changed for the optimal CIFAR-10 models (say the tau and the training epochs, lr and etc)? Appreciate your help very much!