Load Failure around Checkpoint Reading about dcgan-tensorflow HOT 6 OPEN

I fix the same problem by checking this line in model.py:

data = glob(os.path.join("./data", config.dataset, "*.jpg"))

It turned out that my directory name was not compatible with config.dataset. Try to print out the result of this line to make sure that you don't get the data as an empty list. Then celebA can run.

from dcgan-tensorflow.

If you don't have any checkpoint it is failed to load and start training without loading a checkpoint.

I think you're cudnn version is not compatible with the version that TensorFlow uses. You'd better install different cudnn or compile TensorFlow. See tensorflow/tensorflow#1787.

from dcgan-tensorflow.

This is normal when you first training model， and “Load SUCCESS” if exist "checkpoints"

from dcgan-tensorflow.

cuda and cudnn installation must follow https://www.tensorflow.org/versions/master/get_started/os_setup.html#optional-install-cuda-gpus-on-linux， and notice version of cuda and cudnn.

from dcgan-tensorflow.

This is not a problem with Cuda. The same problem exists in my case celeb or your own images. However, when mnist works properly as possible. I am using the CPU so you can exclude the previous presumption.

[inder@dhcppc6 DCGAN-tensorflow] $ python3 main.py --dataset test --is_train True
{'batch_size': 64,
'beta1': 0.5,
'checkpoint_dir': 'checkpoint',
'dataset': 'test',
'epoch': 25,
'image_size': 108,
'is_crop': False,
'is_train': True,
'learning_rate': 0.0002,
'sample_dir': 'samples',
'train_size': inf,
'visualize': False}
[*] Reading checkpoints...
[!] Load failed...

And then nothing happens and the script terminates.

from dcgan-tensorflow.

I have this line instead, still I have load failed...

self.data = glob(os.path.join("./data", self.dataset_name, self.input_fname_pattern))

from dcgan-tensorflow.