@Chainik
Hello,
1)
is the older TensorRT-updating method mentioned here https://www.svp-team.com/forum/viewtopi … 674#p83674
swapping "vstrt.dll and the folder vsmlrt-cuda" only,
still correct and applies for the "v15: latest TensorRT library" from june 2024, together with the newest "RIFE 4.17"?
No other files from the downloaded "vsmlrt-windows-x64-cuda.v15.7z" need to be swapped?
I.e. there is a new folder "vsov" added (100 MB), which is currently not there in "C:\Program Files (x86)\SVP 4\rife"
Using "Rife 4.15 and 4.16" I previously updated this way to the "v14.test4: latest TensorRT and ONNX Runtime libraries" from march 27.
Despite mentioning your performance regression
enjoy" what exactly? 10-15% performance drop
https://www.svp-team.com/forum/viewtopi … 612#p84612
I noticed a drop of 5 % in gpu-utilization and 7 - 8 % in gpu-power, when comparing the same video files for 5 minutes (and had to do all the inferencing for every resolution again).
So I understand it was the correct way to do?
2)
Does "vsmlrt.py" in the "svp/rife" folder, this times has to be swapped too?
It says "vsmlrt.py: Added support for RIFE v4.17 models.", the previous TensorRT version does not mention that.
You stated smt. before, but I have no knowledge if that applies this time too.
only for TRT>=9
with TRT8 updating vsmlrt.py most likely does nothing
[...]