traincascade parameters: precalcValBufSize and precalcIdxBufSize

asked 2014-02-06 07:53:51 -0600

My question is that when I use traincascade, there are these two parameters: -precalcValBufSize and -precalcIdxBufSize.

Can you give me some hints about how to calibrate these numbers? I know that the more memory equals lesser execution time, but how can I set a good balance between the two values when the memory is limited?

Also: how much is too much?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2015-10-02 10:59:50 -0600

Joseph Howse
66 ●4

As described in the official documentation, these parameters are buffer sizes, in Mb. An inspection of the source code shows that the default values are 1024 and 1024. My colleague, Steven Puttemans, recommends setting both parameters to 2048, or even 4096, if you have enough memory available. In other words, traincascade should be able to take advantage of really large buffers.

I do not know how performance is affected by setting one buffer size larger than the other. If you end up running tests, please report your results.