Open mzweilin opened 1 year ago
This PR makes datamodule give input in the [0,1] scale, so that it is compatible with many other existing tools.
transform
untransform
Adversary
preprocessor
Please check all relevant options.
Please describe the tests that you ran to verify your changes. Consider listing any relevant details of your test configuration.
pytest
CUDA_VISIBLE_DEVICES=0 python -m mart experiment=CIFAR10_CNN_Adv trainer=gpu trainer.precision=16
CUDA_VISIBLE_DEVICES=0,1 python -m mart experiment=CIFAR10_CNN_Adv trainer=ddp trainer.precision=16 trainer.devices=2 model.optimizer.lr=0.2 trainer.max_steps=2925 datamodule.ims_per_batch=256 datamodule.world_size=2
pre-commit run -a
Make sure you had fun coding 🙃
@dxoigmn This PR is ready to review again. It should simplify input transforms in our existing datamodules.
What does this PR do?
This PR makes datamodule give input in the [0,1] scale, so that it is compatible with many other existing tools.
transform
anduntransform
inAdversary
so we still use epsilon in integer.preprocessor
in models that were used for transforming input from [0,255].Type of change
Please check all relevant options.
Testing
Please describe the tests that you ran to verify your changes. Consider listing any relevant details of your test configuration.
pytest
CUDA_VISIBLE_DEVICES=0 python -m mart experiment=CIFAR10_CNN_Adv trainer=gpu trainer.precision=16
reports 70% (21 sec/epoch).CUDA_VISIBLE_DEVICES=0,1 python -m mart experiment=CIFAR10_CNN_Adv trainer=ddp trainer.precision=16 trainer.devices=2 model.optimizer.lr=0.2 trainer.max_steps=2925 datamodule.ims_per_batch=256 datamodule.world_size=2
reports 70% (14 sec/epoch).Before submitting
pre-commit run -a
command without errorsDid you have fun?
Make sure you had fun coding 🙃