Daniel Falbel

I suspect that if you set the env var `TORCH_INSTALL=1` before calling `dh-make-R` you will get a working package with all dependencies included - as it will allow downloading the...

HI @egillax , Is the process running torch forked at some point? Usually, forking is not safe when using LibTorch, if doing parallel work it's better to use multi-process parallelization....

This is weird! It's interesting to see some JVM symbols in the backtrace. Do you know where they could come from? Eg: ``` #7 0x00007fece5a97406 in ?? () from /usr/lib/jvm/java-11-openjdk-amd64/lib/server/libjvm.so...

Looking at the traceback to understand what's happening: 1. During autograd, a free memory callback is called and thus the delete tasks event loop start running: ``` #5 0x00007f36286bd604 in...

@egillax ! Thanks for the backtrace. Is the code running into this problem public? I'll try to reproduce it, with a minimal example, but it would be nice to take...

Just to make sure I understand. Should the `available_devices` list contain devices that are allowed by torch, but might not be available at runtime? For instance `'cuda'` might not be...

Cool, OK! I'll add it to the namespace

That would be nice! We have implmented tensorboard support with https://github.com/mlverse/tfevents. Thus if you point tensorboard to a log directory you'll be able to see live training metrics. That's also...

I'm afraid you will need to slighly modify the dataset code to support this. The MNIST dataset code can be found here: https://github.com/mlverse/torchvision/blob/main/R/dataset-mnist.R It seems that you could put the...

I don't think we want to add the parsnip requirement here to this package. Perhaps we can attach parsnip the first time `tabnet` is called, for instance, this is also...