training SSF2020 model with quality 8

[Maryamsana-1998]

Bug

I have been trying to train ssf2020 model with vimeo90k dataset with quality 8 to compare it with the tests reported in tests folder. but i am not getting comparable results for it. Results I'm getting

To Reproduce

1. change the file examples/train_video.py
   net = video_models[args.model](quality=8) and train_dataset.max_frames = 8 test_dataset.max_frames = 8
2. now train this model with command
   python3 examples/train_video.py -d '/data/maryam.sana/vimeo_septuplet/' --cuda

Can you please tell me what should be done to train a quality 8 model correctly? what changes do I have to make?

[fracape]

is your problem specific to that operation point (quality=8)?