With dev dplyr, we now see:
# pak::pak("tidyverse/dplyr")
library(parsnip)
mod <-
linear_reg(engine = 'glmnet', penalty = tune(), mixture = 1) %>%
fit(mpg ~ ., mtcars[1:26,])
multi_predict(mod, mtcars[27:32,], type = "numeric", penalty = seq(0, 1, .2))
#> Warning: Each row in `x` should match at most 1 row in `y`.
#> ℹ Row 1 of `x` matches multiple rows.
#> ℹ If multiple matches are expected, specify `multiple = "all"` in the join call
#> to silence this warning.
#> # A tibble: 6 × 1
#> .pred
#> <list>
#> 1 <tibble [6 × 2]>
#> 2 <tibble [6 × 2]>
#> 3 <tibble [6 × 2]>
#> 4 <tibble [6 × 2]>
#> 5 <tibble [6 × 2]>
#> 6 <tibble [6 × 2]>
Created on 2022-07-21 by the reprex package (v2.0.1)
See tidymodels/tune#526 for initial report and tidymodels/parsnip#772 for one fix.
These warnings, at least, also appear in multi_predict._lognet
in parsnip,
multi_predict._fishnet
in poissonreg, and multi_predict._coxnet
in censored.
This warning might get us in a good few places, but the multi_predict
methods seem like a good place to start looking.
library(tidymodels)
lapply(parsnip:::extensions(), library, character.only = TRUE)
#> [EDIT: output truncated]
methods(multi_predict)
-
multi_predict._C5.0
-
multi_predict._coxnet
-
multi_predict._cubist
-
multi_predict._earth
-
multi_predict._elnet
-
multi_predict._fishnet
-
multi_predict._lgb.Booster
-
multi_predict._lognet
-
multi_predict._mixo_pls
-
multi_predict._mixo_plsda
-
multi_predict._mixo_spls
-
multi_predict._mixo_splsda
-
multi_predict._multnet
-
multi_predict._torch_mlp
-
multi_predict._train.kknn
-
multi_predict._xgb.Booster
-
multi_predict._xrf
-
multi_predict.default
Created on 2022-07-21 by the reprex package (v2.0.1)
PRs have been made where needed.🐙