E tensorflow/stream_executor/cuda/cuda_blas.cc:464] failed to create cublas handle: CUBLAS_STATUS_NOT_INITIALIZED

[mariorioMa]

When I run “ python trainer.py --dataset MNIST --batch_size 36 --num_dis_conv 6 --gan_type wgan-gp”， it shows error as follow：

[PaulZhangIsing]

I tried to download the whole file, and downloaded CIFAR100 dataset.
while I try to run the script:
python3 trainer.py --dataset CIFAR100 --batch_size 25 --num_dis_conv 6 --gan_type wgan-gp
I waited for one day and longer, however, it seems it stucks there and does not progress.

{'id': <tf.Tensor 'shuffle_batch:0' shape=(25,) dtype=string>,
'image': <tf.Tensor 'shuffle_batch:1' shape=(25, 224, 224, 3) dtype=float32>}
2019-10-18 16:18:03.281299: I tensorflow/stream_executor/dso_loader.cc:152] successfully opened CUDA library libcublas.so.9.0 locally
0 10000000 <tf.Variable 'global_step:0' shape=() dtype=int64_ref>
'now is at step 0*******'
[2019-10-18 16:18:04,956] Saved checkpoint at 0

Do you face the same issue if you try to run with CIFAR100 dataset??