error: cuda out of memory

[lucafei]

hi, i met a problem at training DADA network on my own dataset. When i run the train script, it turn out cuda out of memory. Do you know what kind of problem will cause this case? how to solve this? Thank you in advance.

[tuanhungvu]

Hello @lucafei , maybe you can try reduce image resolution and/or batch size.

[shumphries01]

@lucafei were you able to train DADA on your own dataset? I am attempting to do the same and am trying to figure out what changes need to be made. Any help would be appreciated!

[lucafei]

hi, @shumphries01 , i think you can just copy the dataset file like synthia.py and make a little changes to fit your custom dataset. the changes i think is only about get_metadata and getitem which will access to your data. moreover you should create a folder like data to save your training data and youtdatasetname_list in /dada/dataset to put your dataset configuration file similar with synthia_list. finally maybe you need to change some settings in dada.yml in dada/scripts/configs_s2c

[shumphries01]

@lucafei thank you so much for the help! I'll give it a whirl.