Skip to contents

Partition a design into batches that fit in memory

growth_calibration_partition.Rd

Runtime is near-linear in cells but memory is not: per-cell cost rises from about 0.29 MB at 4k cells to 0.34 MB at 39k, and a full structure design runs to hundreds of thousands of cells. Splitting it into batches keeps each run in the linear regime, bounds peak memory, and lets the batches run concurrently or be resumed after an interruption.

Usage

growth_calibration_partition(design, max_cells_per_batch = 12000L)

Arguments

design

A cohort table from growth_structure_design().

max_cells_per_batch

Integer. Cell budget per batch. The default of 12000 corresponds to roughly 4 GB peak, which fits a 16 GB laptop.

Value

design with map_code renumbered within each batch, plus batch, row, and col.

Details

Batches are cut on CELL boundaries, never within a cell, so a mixed cell's cohorts always stay together.

See also

Other growth calibration helpers: extract_landscape_cohort_structures(), growth_add_objective(), growth_auto_window(), growth_best_candidates(), growth_bin_observations(), growth_calibration_design(), growth_calibration_work_root(), growth_climatic_distance(), growth_climatic_weight(), growth_expand_over_pseudo_species(), growth_factorial_ratio_grid(), growth_fitting_windows(), growth_identifiability(), growth_inflation_factor(), growth_mortality_onset_frac(), growth_plot_palette(), growth_pseudo_species_name(), growth_reference_curves(), growth_reference_inflection(), growth_score_fit(), growth_scoring_for(), growth_smooth_observations(), growth_structure_cell_curves(), growth_structure_cohort_table(), growth_structure_design(), growth_structure_summary(), growth_window_for(), plot_growth_calibration(), plot_growth_candidate(), plot_growth_factorial_sensitivity(), plot_growth_structures(), read_growth_scoring(), read_landscape_cohort_structures(), scale_linetype_growth_reference(), write_growth_review_bundle()