mlr3torch
mlr3torch copied to clipboard
Deep learning framework for the mlr3 ecosystem based on torch
Loading mlr3torch breaks `classif.tabpfn` from mlr3extralearners. This works ```r library(mlr3extralearners) Sys.setenv(RETICULATE_PYTHON_ENV = "tabpfn") reticulate::py_discover_config() learner = lrn("classif.tabpfn") learner$train(tsk("sonar")) ``` This fails ```r library(mlr3extralearners) library(mlr3torch) Sys.setenv(RETICULATE_PYTHON_ENV = "tabpfn") reticulate::py_discover_config() learner =...
Not all torch operations are deterministic, even when setting the seed. See the documentation in pytorch: https://docs.pytorch.org/docs/stable/notes/randomness.html pytorch has a way to use reproducible algorithms (at a performance hit), but...
Can you please add some mention of how to use the pt files shown in the example on https://mlr3torch.mlr-org.com/reference/mlr_callback_set.checkpoint.html ? @sebffischer is it possible to start learning in a new...
To make `mlr3torch` flexible to use-cases for which no dedicated task type exist, we should implement a flexible task type that allows to have one more more lazy tensor as...
Is there a way for mlr3torch to make a predefined task for LSTMs? Especially for univariate time series forecasting. I saw something for LSTMs in the original torch book but...