Closed jbmorgado closed 1 year ago
EvoTrees performs training and inference using all threads available to Julia. So there's no special action needed, other than making multiple threads available, which can be achieved in various ways: https://docs.julialang.org/en/v1/manual/multi-threading/.
Notably, either through the JULIA_NUM_THREADS
env variable, or using --threads=8
argument in command line argument.
The documentation is scant about these details, but the Readme talks about training with a 12 thread Ryzen CPU which seems to imply the training process should be multithreaded.
But, when using the examples provided in the documentation, all training is done as single threaded, and, as such, very slow.
Please clarify this in the Documentation and add instruction on how to use multithreading (if it indeed supports it).