amazingren / cit Goto Github PK

Hi,
Links to pre-trained dataset and models no longer work.. could you fix it please?

@Ha0Tang @Amazingren
Hello, Thank you for the amazing repository. I was evaluating the network for JS and other metrics.
I ran into an error while running the test.py for GMM

Can you please help me?

Thank you

Hii @Amazingren
Thank you for your amazing work!
I followed the readme and ran the test. without changing anything, just adding this line of code
torch.backends.cudnn.benchmark = True
Because I got this error: runtimeerror: cudnn error: cudnn_status_success.
The output of stage 2 is so strange .

Could you please tell me what is the problem? and how can I solve it?

Hello, thank you for your amazing work! I tried to train the model, and I finished the GMM step. But when I tried to train the TOM with one GPU(2080Ti), I got a error: "RuntimeError: CUDA error: out of memory " , even if I changed the batch size to 1.
And what I want to ask is : what kind of GPU you used to train ?

python train.py --name TOM --stage TOM --workers 4 --save_count 5000 --shuffle
Namespace(batch_size=4, checkpoint='', checkpoint_dir='/home/admin/workspace/project/CIT/checkpoints', data_list='train_pairs.txt', datamode='train', dataroot='/home/admin/workspace/project/CIT/data', decay_step=100000, display_count=100, fine_height=256, fine_width=192, gpu_ids='0', grid_size=5, keep_step=100000, lr=0.0001, name='TOM', radius=5, save_count=5000, shuffle=True, stage='TOM', tensorboard_dir='/home/admin/workspace/project/CIT/tensorboard', workers=4)
Start to train stage: TOM, named: TOM!

......

RuntimeError: CUDA error: out of memory
terminate called without an active exception
Aborted

Hello,
thank you for your great work!
can you share a colab notebook demo of this amazing repo, it will help everyone to test it and it will be much appreciated
thank you