add butcher methods for nestedmodels

[ashbythorpe]

A simple set of methods for nested_model_fit objects, just iterates through all the inner models and calls axe_() on them.

library(nestedmodels)
library(parsnip)

model <- linear_reg() %>%
  set_engine("lm") %>%
  nested()

nested_data <- tidyr::nest(example_nested_data, data = -id)
fit <- fit(model, z ~ x + y + a + b, nested_data)

weigh(fit)
#> # A tibble: 1,101 × 2
#>    object                  size
#>    <chr>                  <dbl>
#>  1 .model_fit.fit.terms 0.00831
#>  2 .model_fit.fit.terms 0.00831
#>  3 .model_fit.fit.terms 0.00831
#>  4 .model_fit.fit.terms 0.00831
#>  5 .model_fit.fit.terms 0.00831
#>  6 .model_fit.fit.terms 0.00831
#>  7 .model_fit.fit.terms 0.00831
#>  8 .model_fit.fit.terms 0.00831
#>  9 .model_fit.fit.terms 0.00831
#> 10 .model_fit.fit.terms 0.00831
#> # … with 1,091 more rows

res <- butcher(fit)

weigh(res)
#> # A tibble: 1,101 × 2
#>    object                                       size
#>    <chr>                                       <dbl>
#>  1 .model_fit.spec.method.pred.conf_int.post 0.00710
#>  2 .model_fit.spec.method.pred.pred_int.post 0.00710
#>  3 .model_fit.spec.method.pred.conf_int.post 0.00710
#>  4 .model_fit.spec.method.pred.pred_int.post 0.00710
#>  5 .model_fit.spec.method.pred.conf_int.post 0.00710
#>  6 .model_fit.spec.method.pred.pred_int.post 0.00710
#>  7 .model_fit.spec.method.pred.conf_int.post 0.00710
#>  8 .model_fit.spec.method.pred.pred_int.post 0.00710
#>  9 .model_fit.spec.method.pred.conf_int.post 0.00710
#> 10 .model_fit.spec.method.pred.pred_int.post 0.00710
#> # … with 1,091 more rows

data_tst <- data.frame(
  id = 1, id2 = 1, x = 1, y = 1, a = -1, b = 10
)

predict(fit, data_tst)
#> # A tibble: 1 × 1
#>   .pred
#>   <dbl>
#> 1  40.2

predict(res, data_tst)
#> # A tibble: 1 × 1
#>   .pred
#>   <dbl>
#> 1  40.2

[juliasilge]

Thank you so much for this contribution @ashbythorpe! 🙌 I've got a couple of questions about how you set this up.

[juliasilge]

Can you tell me more about how the disabled methods are intended to be found here? I don't seem to find this working for me:

library(butcher)

model <- nestedmodels::nested(
  parsnip::set_engine(parsnip::linear_reg(), "lm")
)
nested_data <- tidyr::nest(nestedmodels::example_nested_data, data = -id)
fit <- parsnip::fit(model, z ~ x + y + a + b, nested_data)

fit$fit$.model_fit <- purrr::map(
  fit$fit$.model_fit,
  axe_fitted,
  verbose = TRUE
)
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`
#> ✔ Memory released: "0 B"

purrr::map(fit$fit$.model_fit, attr, "disabled")
#> [[1]]
#> NULL
#> 
#> [[2]]
#> NULL
#> 
#> [[3]]
#> NULL
#> 
#> [[4]]
#> NULL
#> 
#> [[5]]
#> NULL
#> 
#> [[6]]
#> NULL
#> 
#> [[7]]
#> NULL
#> 
#> [[8]]
#> NULL
#> 
#> [[9]]
#> NULL
#> 
#> [[10]]
#> NULL
#> 
#> [[11]]
#> NULL
#> 
#> [[12]]
#> NULL
#> 
#> [[13]]
#> NULL
#> 
#> [[14]]
#> NULL
#> 
#> [[15]]
#> NULL
#> 
#> [[16]]
#> NULL
#> 
#> [[17]]
#> NULL
#> 
#> [[18]]
#> NULL
#> 
#> [[19]]
#> NULL
#> 
#> [[20]]
#> NULL

^{Created on 2023-03-17 with reprex v2.0.2}

[ashbythorpe]

Apologies, I think I should have been more clear about what I was doing with the verbose option. Since in the "nested_model_fit" object, all the inner models are the same type of model, the functions disabled by each axing function will be the same for every inner model. As a result, my idea was to create a summary of these models, outputting the functions disabled for every inner model and the total memory released, rather than having a separate message for each model. A similar thing is done in ipred.R (except there the disabled methods are already known).

This example demonstrates this (the previous commit should have fixed the issue with the "butcher_disabled" attribute).

library(nestedmodels)
library(parsnip)

model <- linear_reg() %>%
  set_engine("lm") %>%
  nested()

nested_data <- tidyr::nest(example_nested_data, data = -id)
fit <- fit(model, z ~ x + y + a + b, nested_data)

res <- axe_call(fit, verbose = TRUE)
#> ✔ Memory released: "8.47 kB"
#> ✖ Disabled: `print()` and `summary()`

res <- axe_fitted(fit, verbose = TRUE)
#> ✔ Memory released: "5.71 kB"
#> ✖ Disabled: `fitted()` and `summary()`

res <- butcher(fit, verbose = TRUE)
#> ✔ Memory released: "61.96 kB"
#> ✖ Disabled: `print()`, `summary()`, and `fitted()`

# To compare, this is the result of axing one inner model
inner_model <- fit$fit$.model_fit[[5]]$fit

res <- axe_call(inner_model, verbose = TRUE)
#> ✔ Memory released: "920 B"
#> ✖ Disabled: `print()` and `summary()`

res <- axe_fitted(inner_model, verbose = TRUE)
#> ✔ Memory released: "80 B"
#> ✖ Disabled: `fitted()` and `summary()`

res <- butcher(inner_model, verbose = TRUE)
#> ✔ Memory released: "3.92 kB"
#> ✖ Disabled: `print()`, `summary()`, and `fitted()`

Let me know if you think it would be better to show the full output to the user.

[juliasilge]

Yes OK, makes sense to me now, and thank you for fixing the disabled methods! 🙌

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.