Egill Axfjord Fridgeirsson

different GPU memory use between torch in R and pytorch

10

I have some code [here](https://github.com/egillax/torchComparison) where I'm running the same models on the same random data to compare torch in R vs pytorch. One thing that I'm noticing is that...

Is it possible to detect if tensor is sparse or dense?

4

I'm trying to make a custom collate_fn where some of my data is sparse and some dense. With the sparse data I want to stack and then use to_dense() but...

Allow sparse matrix input to tabnet

1

Currently only dense matrices and dataframes are supported for the x matrix. This limits on how large data can be used with the current implementation. It would be good to...

Similar to [pytorch](https://pytorch.org/docs/stable/generated/torch.cuda.empty_cache.html) This would be useful for example when running hyperparameter tuning where you train a model multiple times.

add is_sparse method for tensors

1

solves #736 Full credit goes to @rdinnager since I got this from his branch [here](https://github.com/rdinnager/torch/tree/feature/autograd_functional). I just needed this functionality again and there's been no movement on that branch since...

deadlock when training torch model

8

Hi @dfalbel , I'm having an issue that the model training just hangs when training models. This doesn't happen always but has now happened twice in the last week. I'm...

This can speed up some operation for newer generations of GPUs: https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices

[WIP] New external validation function

3

For discussion @jreps @lhjohn I made a new function [externalValidate](https://github.com/OHDSI/PatientLevelPrediction/blob/external_validation_strategus/R/ExternalValidatePlp.R#L338) to simplify external validation for our network study and in general in Strategus. This function allows us with one call...

Small quality of life change. So when you don't want to restrict in any way you don't have to create a `restrictPlpDataSettings` to add it as an argument to `getPlpData`

I think it would be good to add time measurement to each step in runPlp and the total time. I already did so in the arrow branch when I was...