Comments (7)
DavisVaughan
commented on September 26, 2024
A reference to this article is mentioned in one sentence at the bottom of the docs for step_embed()! https://keras.rstudio.com/articles/faq.html#how-can-i-obtain-reproducible-results-using-keras-during-development
from embed.
topepo
commented on September 26, 2024
Can you provide a reprex where you run the same recipe more than once? Try setting the seed before prepping the recipe. The underlying code should set the TF seed(s) based on R's random number stream.
from embed.
ciberger
commented on September 26, 2024
Thanks for your quick replies guys!
Basically, each time I rerun the embeddings I get a new session seed number, it seems there is no way to modify the seeds param on the tf_coefs2 function.
On the other hand, using use_session_with_seed(...) before calling the step_embed function does not work either.
Test 1
library(vcd)
data(Arthritis)
require(recipes)
require(embed)
#> Loading required package: embed
recipe.obj <- recipe(formula("Improved ~ Treatment + Sex + Age"), data = Arthritis) %>%
step_embed(
all_nominal(), -all_outcomes(),
predictors = vars(all_numeric()),
outcome = vars(Improved),
num_terms = 2,
hidden_units = 0,
options = embed_control(
epochs = 400, validation_split = .2
)
) %>%
prep(training = Arthritis)
#> Set session seed to 1695 (disabled GPU, CPU parallelism)Created on 2019-01-20 by the reprex package (v0.2.1)
Test 2
library(vcd)
data(Arthritis)
require(recipes)
require(embed)
#> Loading required package: embed
recipe.obj <- recipe(formula("Improved ~ Treatment + Sex + Age"), data = Arthritis) %>%
step_embed(
all_nominal(), -all_outcomes(),
predictors = vars(all_numeric()),
outcome = vars(Improved),
num_terms = 2,
hidden_units = 0,
options = embed_control(
epochs = 400, validation_split = .2
)
) %>%
prep(training = Arthritis)
#> Set session seed to 5576 (disabled GPU, CPU parallelism)Created on 2019-01-20 by the reprex package (v0.2.1)
from embed.
skeydan
commented on September 26, 2024
To me it seems that when
tf_coefs2 <- function(x, y, z, opt, num, lab, h, seeds = sample.int(10000, 4), ...) {
gets called, you get different seeds vectors in fresh R sessions, as well as when calling the function several times in a row:
# restart R
sample.int(10000, 4)
sample.int(10000, 4)
## restart R
sample.int(10000, 4)
sample.int(10000, 4)
> # restart R
> sample.int(10000, 4)
[1] 8528 4771 4492 7039
> sample.int(10000, 4)
[1] 2927 7478 803 7227
Restarting R session...
> # restart R
> sample.int(10000, 4)
[1] 4363 6519 7028 6792
> sample.int(10000, 4)
[1] 7621 7222 9856 2809
from embed.
topepo
commented on September 26, 2024
If have never set the seed, you should get different results.
Since the recipe steps pull the tensorflow seeds from R's random numbers, you only need to set R's seed to get the same TF seeds:
library(embed)
#> Loading required package: recipes
#> Loading required package: dplyr
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
#>
#> Attaching package: 'recipes'
#> The following object is masked from 'package:stats':
#>
#> step
data(okc)
set.seed(34523)
take_1 <-
recipe(Class ~ age + location, data = okc) %>%
step_embed(location, outcome = vars(Class),
options = embed_control(epochs = 10)) %>%
prep(training = okc) %>%
tidy(number = 1)
#> Set session seed to 5504 (disabled GPU, CPU parallelism)
set.seed(34523)
take_2 <-
recipe(Class ~ age + location, data = okc) %>%
step_embed(location, outcome = vars(Class),
options = embed_control(epochs = 10)) %>%
prep(training = okc) %>%
tidy(number = 1)
#> Set session seed to 5504 (disabled GPU, CPU parallelism)
all.equal(take_1, take_2)
#> [1] TRUECreated on 2019-01-21 by the reprex package (v0.2.1)
from embed.
skeydan
commented on September 26, 2024
That makes sense!
from embed.
github-actions
commented on September 26, 2024
This issue 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.
from embed.
Related Issues (20)
- Cannot `update()` a tuneable `step_umap()` HOT 2
- Create groupings in the reference of pkgdown HOT 1
- Poisson models fail for likelihood encodings HOT 2
- Release embed 0.2.0 HOT 1
- FR: For each of the UMAP clusters, information/ID on values (from which columns) assigned to which UMAP clusters would be nice HOT 6
- step_umap crashing Rstudio HOT 18
- catboost method to embed categorical variables HOT 11
- Release embed 1.0.0 HOT 1
- step_woe errors uninformatively if outcome isn't a factor HOT 2
- Allow step_collapse_stringdist to accept different distance methods HOT 2
- Metrice argument for step_umap function HOT 2
- Custom metric for step_umap HOT 2
- Upkeep for embed HOT 1
- remove tidyr_new_interface() check HOT 1
- Test that all tunable.step_*() are specified correctly HOT 1
- Use rlang errors HOT 1
- step_embed() should have `keep_original_cols` argument HOT 1
- Release embed 1.1.0 HOT 1
- Add missing infrastructure tests HOT 1
- Release embed 1.1.1 HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from embed.