search

mit-han-lab / data-efficient-gans

[NeurIPS 2020] Differentiable Augmentation for Data-Efficient GAN Training
https://arxiv.org/abs/2006.10738
BSD 2-Clause "Simplified" License
1.28k stars 174 forks source link

policy=“translation” is failed #3

Closed Johnson-yue closed 4 years ago

Johnson-yue commented 4 years ago

Hi， I test you code with two framework, and two dataformat, and the same result is "translation" policy is failed!

Envs: Python: 3.7 Ubuntu: 1804 CUDA 10.1 Pytorch: 1.4.0

Test1:

I use your code in BMSG-GAN repo and train 3-channel dataset(102 -flower) but only using 755 samples for checking, when the policy included "translation" , and failed to train.

RuntimeError: CuDNN error: CUDNN_STATUS_NOT_SUPPORTED. This error may appear if you passed in a non-contiguous input

but if policy I just used "color,cut", it work fine!

Test2:

In my own GAN with 1-channel dataset, the performance is the same, if not use "translation" policy it work fine, otherwise it failed

Johnson-yue commented 4 years ago

When I use Python: 3.5 this bug is disappeared!!!

zsyzzsoft commented 4 years ago

I'm trying to fix it, but I cannot replicate the error. Could you help me to verify whether the updated code works now on Python 3.7? Thank you!

junyanz commented 4 years ago

maybe we should mention PyTorch/Tensorflow version in the README.

zsyzzsoft commented 4 years ago

The DiffAugment operations are supposed to work on all common PyTorch/TensorFlow versions. If it doesn't work properly on some version, just let me know!

Johnson-yue commented 4 years ago

@zsyzzsoft this code work fine on Python 3.5， but not work on Python 3.7， Did you try on Python 3.7??

zsyzzsoft commented 4 years ago

I'm using Python 3.7.6. Did you try the updated code? Could you share the complete error message, including which line that yields the error?

Johnson-yue commented 4 years ago

ok， if I occur same problem， I will report full log here ， and I just close this issue temporarily

jjparkcv commented 3 years ago

Has anyone solved this issue? I'm seeing this CuDNN error using Python 3.6.8.

jjparkcv commented 3 years ago

Has anyone solved this issue? I'm seeing this CuDNN error using Python 3.6.8.

Fixed the issue by calling .contiguous() at the end of the rand_translation() function.

zsyzzsoft commented 3 years ago

Thanks! Fixed in master.