Rife v4.14 lite about practical-rife

[WolframRhodium]

These warnings are not related to either accuracy or performance. They simply say that TensorRT cannot find previous profiling records (because these operations are new) and it need to start profiling later.

[banjaminicc]

This issue should be on vs-mlrt. Not practical RIFE.

This should only be relevant if you used the provided inference framework here instead of vs-mrlt as it would be unclear whether the bandwidth limitation comes from TensorRT or Pytorch or different codes

[netExtra]

Wolfram suggested I put it in the authors page Is this not the authors page?

[banjaminicc]

Ah, I just think that if you want this to be more accurate it would be better to test the speeds using the implementation that hzwer provided

[netExtra]

Ah, I just think that if you want this to be more accurate it would be better to test the speeds using the implementation that hzwer provided

I just want to understand the errors. It doesn't affect the result but I want to have some idea of what they mean.

[netExtra]

These warnings are not related to either accuracy or performance. They simply say that TensorRT cannot find previous profiling records (because these operations are new) and it need to start profiling later.

OK so all of this is just part of the profiling process and since these records haven't been generated yet, they are kind of like cache misses.