How to define the percentage of training data used in segmentation tasks

[SZUHvern]

I would like to know how to define the percentage of training data used in segmentation tasks? I couldn't find the relevant variables in the code.

Thanks again for sharing the code.

[DopamineLcy]

Please use the config https://github.com/RL4M/MRM-pytorch/blob/3730ba49a678eda870284205178bfe25627aaed8/Siim_Segmentation/configs/mrm/upernet_mrm-base_fp16_8x2_512x512_160k_siim_10per.py

[SZUHvern]

Please use the config https://github.com/RL4M/MRM-pytorch/blob/3730ba49a678eda870284205178bfe25627aaed8/Siim_Segmentation/configs/mrm/upernet_mrm-base_fp16_8x2_512x512_160k_siim_10per.py

Sorry, do you mean 'max_iters' here is the number of training dataset?

MRM-pytorch/Siim_Segmentation/configs/_base_/schedules/schedule_5k.py

Line 7 in 3730ba4

runner = dict(type='IterBasedRunner', max_iters=5000)

[DopamineLcy]

You could specify the config in https://github.com/RL4M/MRM-pytorch/blob/main/Siim_Segmentation/ft.sh
If you want to use 10% of the training set data, you may set the config configs/mrm/upernet_mrm-base_fp16_8x2_512x512_160k_siim_10per.py.
And If you want to use 100% of the training set data, you may set the config configs/mrm/upernet_mrm-base_fp16_8x2_512x512_160k_siim.py, which is set by default.