Open 649459021 opened 5 months ago
Dear author, Thank you for releasing this great work.
When I reproduce the training process, I find that the metrics on the real dataset are getting worse after 100k, but getting better on ImageNet-Test.
What do you think is causing this? Is there any way to mitigate this?
Looking forward to your reply.
Dear author, Thank you for releasing this great work.
When I reproduce the training process, I find that the metrics on the real dataset are getting worse after 100k, but getting better on ImageNet-Test.
What do you think is causing this? Is there any way to mitigate this?
Looking forward to your reply.