Laurae

Newer LightGBM GPU results using not all threads, restricting to 1 NUMA node and physical cores, excluding histogram building time (negligible, 0.04x seconds for 0.1m,, 0.1xx seconds for 1m, 1.180s...

@trivialfis We are not passing any watchlist, only the objective for gradient / hessian is computed.

CUDA LightGBM: some results from myself here: https://gist.github.com/Laurae2/7195cebe65887907a06e9118a3ec7f96 (**VERY experimental**) Using commit Microsoft/LightGBM@df37bce (25 Sept 2020). GPU usage increases as lower number of GPU is used (ex: 80% for 1...

@szilard `max_leaves = 1024` builds very deep trees. With 32 leaves it should already build deeper trees than `max_depth = 10` in most cases. `max_depth` was available before LightGBM GPU...

@szilard Try this: https://github.com/Microsoft/LightGBM/blob/master/docs/Key-Events.md

Hardware/Software: https://github.com/szilard/GBM-perf/issues/12 hist xgboost, 1 model: | Size | Time (1T) | Time (9T) | Time (18T) | Time (36T) | Time (70T) | | ---: | ---: | ---:...

Dumb "tuning" abusing knowledge of only 1 feature is relevant allows us to choose a proper more maximum depth of 2 instead of an absurd 10, which leads us out...

Assuming the sample strategy is the same (and using probs models), it is possible to combine train models into a caretList. The creation of indexes must be stored into a...

@ogrisel You can apply for a free Intel VTune license for profiling your code if you do research. It will be much better than the numba profiler.

@stuartarchibald You can use the numba profiler here: https://github.com/numba/data_profiler (it just adds the signatures in reality). Incurs overhead penalty. Still better to use Intel VTune for real profiling though (way...