search

songzijiang / LGAN

Source codes for LGAN
Apache License 2.0
8 stars 0 forks source link

About patch size #1

Closed Rax111 closed 1 year ago

Rax111 commented 1 year ago

Hello, what is in the yml file What are the patch sizes of ×2, ×3 and ×4 respectively?

songzijiang commented 1 year ago

Thank you for your attention to our work! configs/config.yml is the configuration of the network. There are no rules about how to set patch sizes. You can set 'patch size' considering your GPUs` Memory. From experience, setting 'patch size' between 60 and 120 is a good choice. If you have enough Memory, you can set a larger path size for better perfencement.

The .yml file has been updated recently, I will upload a new config.yml later. Hope this can help you.

Rax111 commented 1 year ago

Thank you so much

songzijiang commented 1 year ago

I have checked the config.yml. The latest version is the correct configuration. Using the default configuration in config.yml will train an X4 model for 1000 epochs. I will update the meanings of the parameters, you can check the config.yml later.