-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Pipeline-compact #338
Pipeline-compact #338
Conversation
sigmafelix
commented
Jun 11, 2024
•
edited
Loading
edited
- V 0.3.0
- Compact pipeline: feature calculation is tested. Also implemented base learners, but yet untested.
- Function restructuring: amadeus functions are removed, meta learner functions are kept for reference
- TODO: Documentation
- TODO: tests
- Tests were removed accordingly
- aadt is replaced by nei
- Removed (a-la-)duplicate functions of amadeus - Rd cleaned - stdt dependency remains; soon to be deprecated - DESCRIPTION update: amadeus is listed in Remotes:
- mostly skeletonized until meta learner fitting - TODO: prediction grid covariate computation, status check, etc.
- Meta learner fitting - prediction grid - documentation for pipeline functions
- Base structure is established - To be filled and refactored then be tested
- mermaid chart - package name change in _targets.R: scomps to chopin
- gets adam initialization error - FIXME: the returned tensor has 17 instead of 18 in the paper example
- running (frozen)
- use non-function as function error
- Pipeline main code is split into several files for maintenance - MLP with brulee + tidymodels: dev
- calculation code (customized for geos) - gmted is on hold - Stale codes were removed
- 10 locations on a single day - run okay except for four targets (TODO) in feature calculation
- 10 points, single day: each calculation flow works as expected - temporal field transformation/renaming is required to join data.frames successfully - TODO: scaled example (full data + several months) - TODO: looking at possibility of tar_make_future() deprecation - TODO: shift to mirai for crew compatibility?
- does crew(mirai)+future work?
- TODO: clean space-time feature data.frames for neater layout - FIXME: join tables for space-time features (ie. heterogenous temporal resolution)
- Added family tag in roxygen2 doc - Example errors fixed
- spatiotemporal cv based on anticlust::balanced_clustering - Adjacent initial clusters are paired into predefined number of secondary sets
- generate_cv_index: generation mode and preprocessing - vis_rset: spt visualization
- Family reclassified - Lint
- no whitespace in keywords
- error-on option is changed to "error" to avoid remote package import errors - Example naming error fix
@kyle-messier I managed to pass the PR checks by putting most of the new functions into nocov status (for the time being, of course) and changing check-standard.yaml setting to fail only if R CMD check returns an error. I will improve the coverage in the next PR. Thank you. |
- "under construction" flag removed - _targets.R and pipeline_base_functions.R in inst/targets are synchronized
_targets.R
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Any issue with user specific paths here? "/ddn/gs1/home/songi2/r-libs"
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've been using a user library path for additional packages that are not available in the system default library path. I could make a subdirectory in the team project directory or somewhere else then will change the value in _targets.R
. I think everyone in the DDN user group grpkpmessier
will be able to read the library in /ddn/gs1/home/songi2/r-libs
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Each review was commented.
_targets.R
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've been using a user library path for additional packages that are not available in the system default library path. I could make a subdirectory in the team project directory or somewhere else then will change the value in _targets.R
. I think everyone in the DDN user group grpkpmessier
will be able to read the library in /ddn/gs1/home/songi2/r-libs
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@sigmafelix Just a few things I noticed as comments.
input/tidymodels_stcv_integration.R
Outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are L10-12 deprecated or still in use?
source("R/manipulate_spacetime_data.R")
source("R/cross_validation.R")
source("input/Rinput/processing_functions/filter_unique_sites.R")
`
I am not seeing R/cross_validation.R
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This file is no longer relevant, so I removed the file.
tools/generate_test_data.Rmd
Outdated
@@ -21,7 +21,9 @@ knitr::opts_chunk$set( | |||
) | |||
if (!require(NRTAPmodel)) { | |||
pak::pak("Spatiotemporal-Exposures-and-Toxicology/NRTAPmodel@isong-calc-covars") | |||
pak::pak("Spatiotemporal-Exposures-and-Toxicology/amadeus") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Old username GitHub paths?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Fixed the path and cleaned redundant lines.
- Old or irrelevant files removed - Fixed GitHub usernames - TODO: remove tools entirely
@kyle-messier Thank you very much for reviewing my PR thoroughly. I think I reflected the comments into the recent push and I will merge it into |