Some questions about the IN1K-Self model

[zh133767710]

Hello, I am trying to replicate the results of SLCA method. I have some questions I'd like answered.

The IN1K-Self model mentioned in this paper seems to be at variance with the mocov3 model published in the code, since in my replication the results using the mocov3 model provided in the code are almost indistinct from 21k.

I kindly ask you to provide relevant links to IN1K-Self's model for your experiments. This would greatly help me to accurately replicate the experiments in your work.

Thank you for your time and attention. Looking forward to your reply.

[GengDavid]

Hi, @zh133767710. Thanks for your interest and efforts. The IN1K-Self model used in our experiments is the one provided in the README, which is extracted from the official MoCoV3 checkpoint.

I am not sure why you are getting the same results as the IN-21k pre-train. Do you have any training log about your training?

[zh133767710]

Thanks for sharing, I will check my code again.