Context
Following up on this discussion and the addition of benchmarks in issue #106.
Description
Consider splitting out GitHub Actions such that one action is not building
- 5 serotypes (denv1, denv2, denv3, denv4, all)
- × 2 genetic regions (full genome, E gene)
- = 10 phylogenetic trees
Similar to how GitHub Actions is split by build in WNV, consider splitting this into separate phylogenetic_X jobs, perhaps by serotype. This would allow each job to use 4 cores without overloading a single runner.
An added benefit is that we could independently trigger the rebuild of a serotype tree only when new sequences for that serotype are available. Currently, new sequences—regardless of serotype—trigger the rebuilding of all 10 trees which is currently taking ~4hrs.
Perhaps add a config to use a reduced test set or subsample, during PR testing.
Context
Following up on this discussion and the addition of benchmarks in issue #106.
Description
Consider splitting out GitHub Actions such that one action is not building
Similar to how GitHub Actions is split by build in WNV, consider splitting this into separate phylogenetic_X jobs, perhaps by serotype. This would allow each job to use 4 cores without overloading a single runner.
An added benefit is that we could independently trigger the rebuild of a serotype tree only when new sequences for that serotype are available. Currently, new sequences—regardless of serotype—trigger the rebuilding of all 10 trees which is currently taking ~4hrs.
Perhaps add a config to use a reduced test set or subsample, during PR testing.