Sebastian Fischer

[feature request] Way to load state dict for corrup model (invalid external pointers)

1

Hey, thank you very much for taking the time! I have some questions regarding the serialization of torch models. In mlr3torch we have learners (R6 classes) that have a field...

Request: provide `to(device = "cuda_pseudo")` to test that code runs on GPU

1

First of all, thank you for your great package! In order to test that code runs on CPUs and GPUs on machines (especially for CI) that does not have a...

Possibility to adjust minimal distance to the bottom

1

First of all, thank you for radian, it is a great console! I am finding myself with an issue however when working with Nvim-R + radian. When I am on...

as_task_xxx should enable setting weights ect

1

I think making this work would be a good user-friendly change, what do you think? ```r iris$weight = runif(150) task = as_task_classif(iris, target = "Species", weight = "weight") ```

Oftentimes I write `msr("classif.acc")` where it should be enough to just write `"classif.acc"` ``` rr = resample(tsk("iris"), lrn("classif.rpart"), rsmp("holdout")) rr$score("classif.acc") ```

BREAKING CHANGE: Introduce default_params

5

* Improves transparency: Learners should set those values in the initialize function (and not e.g. in $.train) * Could be used in description for automatic display of defaults * Could...

Give learners "tags" like in mlr3pipelines

2

I think it would be a good addition to tag learners according to the category to which they belong. E.g. in `mlr3extralearners` we have many different implementations of very similar...

I suggest that Learner should have a private field `.default_params` where the default parameters are stored. This has the advantage that: 1. When calling `$reset()` the parameters can also be...

Extend version mismatch warning to `$packages`

1

Why do we only compare the mlr3 version of the learner with the current version when calling`$predict()` ?