Jerry Ling
Jerry Ling
I'm wondering if you just need `GPUArrays.jl` for interface
yeah, thanks for the GPU support, I'm just experience slow loading time on my poor cluster login node so no rush
what about https://github.com/JuliaGPU/GPUArrays.jl/tree/master/lib/GPUArraysCore
>It's nice to have GPU support but personally usually I do not have access to dedicated GPUs, so to me it rather seems yeah I agree with this
 :(
yeah I think unfortunately it's not part of the model parameter `num_rounds` is the maximum rounds in some sense, early_stopping_rounds is the number of rounds without improvement for the fit...
btw related question, is it possible to error when unsupported keywords gets passed to XGBoost? for example the `early_stopping_rounds`, and because it didn't error, I keep worrying about if I...
yeah, well, we have to do this externally in Julia because we can't pass a callable to `libxgboost` for evaluation set evals
actually >it's bad enough that we are already trying to do this to identify warning messages, I'm not sure if it's for or against the idea to parsing logs to...
one reason to not do it in Julia would be performance, also keep in mind we'd have to some how evaluate it with the same metric as the training and...