In jbryer/mldash: Evaluating Machine Learning Models Across Many Datasets

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

You can contribute to this project by adding any one (or all) of three components: datasets; models; model runs. By adding these components, other testers will be able to run, validate, compare and build upon your contributions.

Contributing requires the following steps: * Creating a Github fork of the mldash project (under your own username) * Locally git-cloning your fork * Creating a new file for your dataset, model or model-run * Git-adding and git-commiting your new file * Pushing your changes back to your Github fork * Creating a pull request to the original mldash project

To see examples of each component, you can go to the mldash/inst folder.

Prerequisites

You will need to complete typical Github pre-tasks: registration; creating an personal access token; creating a fork; and finally, creating a clone of your Github fork on your local workstation.

After creating a Github account, go to your profile settings and select 'Developer Settings' and then 'Personal Access Tokens' and 'Tokens (classic).' After following the on screen directions to generate a new token, be sure to save your token in a safe place; you will need it later.

Next, create your own fork of the mldash project at https://github.com/jbryer/mldash and click on the 'fork' button so you can keep a downstream copy of the project. While on the page of your forked repository, retrieve SSH URL link of your project.

In your local workstation, perform a git clone with the URL link from the previous step. If you issue a 'git remote show origin' command in a command line terminal, you should see something similar to the following:

$ git remote show origin
* remote origin
  Fetch URL: https://github.com/cliftonleesps/mldash
  Push  URL: https://github.com/cliftonleesps/mldash
  HEAD branch: master
  Remote branches:
    master     tracked
    ml_results tracked
  Local branch configured for 'git pull':
    master merges with remote master
  Local ref configured for 'git push':
    master pushes to master (fast-forwardable)

Contributing a New Dataset

To add a dataset, simply define it in Debian Control Format as described here. After creating a new *.dcf file, use git to add, commit and push your new file to Github (substitute NEW_DATASET with the name of your particular file):

$ git add inst/datasets/NEW_DATASET.dcf
$ git commit -m "Adding new dataset" inst/datasets/NEW_DATASET.dcf
$ git push

After a successful git push action, go to Github and confirm your new dataset appears in your forked repository.

Your last task is to click on the 'Pull requests' tab and then "New pull request" and "Create pull request" button.

At this point, a project maintainer in the upstream mldash project will review your new *.dcf file and approved of the merge.

Contributing a New Model

Similarly, you can go to the Creating New Models page for details on creating a new model. Once you have a new model DCF file, you follow the same steps when adding a new dataset file (substitute NEW_MODEL with your model's name):

$ git add inst/models/NEW_MODEL.dcf
$ git commit -m "Adding new model" inst/models/NEW_MODEL.dcf
$ git push

As with adding a new dataset, you then go to Github and create a pull request against the https://github.com/jbryer/mldash project.

Contributing a Model Run

After running models mldash::run_models() and saving the results in a dataframe (see below), you can save your results to a file in /inst/results file (substitute RESULTS_FILE with your filename), inside of your R session:

ml_results <- mldash::run_models(datasets = ml_datasets, 
                                 models = ml_models, 
                                 seed = 2112)

saveRDS(ml_results, file.path("inst","results","RESULTS_FILE.rds"))

You now have a new results file in your mldash/inst/results directory and in your terminal, you can now commit and push it to Github:

git add inst/results/RESULTS_FILE.rds
git commit -m "Contributing a model run" inst/results/RESULTS_FILE.rds
git push

Again, the last step is going to Github and creating a pull request against the https://github.com/jbryer/mldash project.

jbryer/mldash documentation built on March 4, 2023, 9:35 p.m.