gsgxnet

Yeah, PyTorch within Carnet (or Carnet plus) would be great. There is some IOS support in PyTorch now: https://github.com/pytorch/ios-demo-app It seems to be a reduced functionality version. So might not...

To prepare for taking part in the R TabNet Online Meetup in Brussels I did setup my environments for both the R and Python TabNet. Sadly the performance differences are...

Hi @cregouby, thanks for looking further into it. I am trying to figure out whats going on myself, but did not make any real progress. What I have: ------- **PyTorch**...

Using `profvis` for the `tabnet_fit`: ``` profvis({ md

I have a second remark to my profiling of the code. If that `.Call` with 87.330 sec is really the time spent in `libtorch`, mostly GPU leaning, that would also...

> Thanks a lot @gsgxnet for this investigation, very helpfull. May I propose to split that thread in as many topics as needed ? > > * [Dataloader cannot use...

I wonder, is this an OS set limit? Does Sonoma (macOS 14) change anything?

> It is enforced by the OS. There have been somewhat successful attempts to change the memory allocation with a custom kernel/module but they come at a price; significantly reduced...

At this time there is no CatBoost 1.2.3 available at conda-forge. ``` conda search catboost | grep '1.2.[23]' catboost 1.2.2 py310hbe9552e_0 conda-forge catboost 1.2.2 py311h267d04e_0 conda-forge catboost 1.2.2 py38h10201cd_0 conda-forge...

Same here, would be great if we could config several upstream hosts for this add-on