search

deepak112 / Keras-SRGAN

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network implemented in Keras
276 stars 146 forks source link

Memory Error when training the model #35

Open mustaeenqazi opened 4 years ago

mustaeenqazi commented 4 years ago

Hi Deepak, Thanks for the amazing code, a small issue, when i start training the model with your simplified version of code, the Ram gets exhausted. although i have 12GB ram on my google colab, I tried reducing the size of images but no results

hongyanghu commented 4 years ago

have you solve the problem?pls

rcbrigg commented 4 years ago

I had the same problem. What batch size are you using? Reducing that should help. I also halved the number of neurons in the dense layers and number of filters in the conv layers, which of course will impact performance.

mustaeenqazi commented 4 years ago

The problem is solved by using small batch size and lower number of images. but you can also try to upgrade your memory.

BassantTolba1234 commented 3 years ago

Dear Sir, Brilliant work , thank you for sharing your code.. Please, according to your GPU version what the time did you take to finish training your network ?? I'm waiting for your fast reply..

BassantTolba1234 commented 3 years ago

Dear Sir, Brilliant work , thank you for sharing your code.. Please, according to your GPU version what the time did you take to finish training your network ?? I'm waiting for your fast reply..

BassantTolba1234 commented 3 years ago

Please all, I need the code of implementation this part the part is

{The SRResNet networks were trained with a learning rate of 10−4 and 106 update iterations. We employed the trained MSE-based SRResNet network as initialization for the generator when training the actual GAN to avoid undesired local optima.{

MuterDear commented 2 years ago

Could you please tell me which size of your image before super resolution?