search

tarun005 / FLAVR

Code for FLAVR: A fast and efficient frame interpolation technique.
Apache License 2.0
463 stars 70 forks source link

training issue about PSNR #47

Closed ss00atbupt closed 8 months ago

ss00atbupt commented 2 years ago

Hi, tarun ,excellent work on video interpolation! I tried run your code , but I have some trouble. I set my config as batch_size=2, beta1=0.9, beta2=0.99, checkpoint_dir='.', cuda=True, data_root='/home/Changchen/dataset./vimeo_septuplet', dataset='vimeo90K', exp_name='exp', joinType='concat', load_from=None, log_iter=60, loss='1*L1', lr=0.0002, max_epoch=50, model='unet_18', n_outputs=1, nbr_frame=4, nbr_width=1, num_gpu=1, num_workers=16, pretrained=None, random_seed=12345, resume=False, resume_exp=None, start_epoch=0, test_batch_size=1, upmode='transpose', use_tensorboard=False, val_freq=1 At the beginning, psnr was normal about 20,but it has gradually decreased to about 14. I wonder why it seems to be misconvergence. Thank you for any help!

ss00atbupt commented 2 years ago

By the way,I train my modle on vimeo-90k.

ss00atbupt commented 2 years ago

Hi,@tarun005 ! I tried to set lr as 0.002,now it seems better than before.My model's psnr has raised a little! Should I set lr bigger than 0.002 at the beginning? Thank you for any possible help!

tarun005 commented 2 years ago

Can you make sure that the versions of the PyTorch are the same as recommended? I did not face any such issues at my end.

ss00atbupt commented 2 years ago

Thanks for your reply! I didn't use the same PyTorch version .Somehow I can't use it , CUDA version can't match when I use PyTorch 10.1. So I used 10.2 . Does the version affects the training ?