review - Githubissues

mlr-org / mlr3keras

Deep learning for mlr3

GNU Lesser General Public License v3.0

36 stars 3 forks source link

review #53

Open be-marc opened 2 years ago

be-marc commented 2 years ago

Application of mlr3keras from a user perspective. Work in progress.

[ ] README.md - Links to learners do not work.
[ ] README.md - reticulate::conda_create() is called with python=3.8 but keras::install_keras() switches to python version 3.7
[ ] Why does LearnerClassifKerasFF$new() starts installing keras and tensorflow again?

Installation that works for me:

reticulate::install_miniconda()

reticulate::conda_create(
  envname = "mlr3keras",
  packages = c("pandas", "python=3.7")
)

Sys.setenv(RETICULATE_PYTHON = "/home/marc/.local/share/r-miniconda/envs/mlr3keras/bin/python")
keras::install_keras("conda", tensorflow="2.3.1", envname="mlr3keras")
reticulate::use_condaenv("mlr3keras")

[ ] If reticulate is not forced to use /home/marc/.local/share/r-miniconda/envs/mlr3keras/bin/python, it uses /home/marc/.local/share/r-miniconda/envs/r-reticulate/bin/python with python 3.6. Calling reticulate::use_condaenv("mlr3keras") does not change this behavior.
[ ] README.md - Do not use lrn as a variable name