hzwer

As far as I know, many people are using this version： https://github.com/hzwer/arXiv2020-RIFE/issues/41#issuecomment-785548531 Maybe it’s helpful to check the comment of flowframes https://nmkd.itch.io/flowframes

v2.2 https://drive.google.com/file/d/1luM3AnlxR-pEH4ugOA8iHMBgYYPONxOj/view?usp=sharing v2.3 https://drive.google.com/file/d/1VdzN9beh7qi4up9hS7dsEg6WOwUdBh-Q/view?usp=sharing v2.4 https://drive.google.com/file/d/1wsQIhHZ3Eg4_AfCXItFKqqyDMB4NS0Yd/view?usp=sharing

@dcominottim Hi, we are not currently support fp16.

Your suggestion is sound, I can give it a try.

In fact, the v4.0 version already contains the data of anime10K. I recently tried developing the "anime" and "real world" versions separately, but found no improvement.

v4.4 model is tuned for anime. We will further optimize it.

v2.3 model is unexpectedly good. v4.0 achieves both effects and performance; according to the latest large-scale evaluation report, v4.0 should be a suitable choice

I am developing a new algorithm that can be used as a supplement to any video frame synthesis algorithm. For example, it can be used to repair bad frames generated...

I meet the same issue using cuda11.0. Fortunately, I re-used cuda10.2 + cudnn8.0.4 + torch1.7.1 and successfully ran the demo.py provided by the author.

> > I meet the same issue using cuda11.0. Fortunately, I re-used cuda10.2 + cudnn8.0.4 + torch1.7.1 and successfully ran the demo.py provided by the author. > > Which version...