You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
So, we need to set s=0.5 for CIFAR-10 experiments if we obtain the experimental results more accurately.
Best regards,
The text was updated successfully, but these errors were encountered:
nzw0301
changed the title
Slightly difference strength parameter in CIFAR-10 setting
Slightly different strength parameter in CIFAR-10 setting
Aug 12, 2020
Hi, thank you for sharing SimCLRv1 code by using pytorch. This repo is really helpful.
I realised that the strength parameter
s
ofColorJitter
is slightly different from the paper's setting.Appendix B.9.
CIFAR-10
of SimCLRv1 paper saysSo, we need to set
s=0.5
for CIFAR-10 experiments if we obtain the experimental results more accurately.Best regards,
The text was updated successfully, but these errors were encountered: