Fit a BatchBirdFlowTrainer

Fits one or more BirdFlow models determined by the trainer's parameter grid, submitting jobs via batchtools and retrying failed or expired jobs.

Usage

# S3 method for class 'BatchBirdFlowTrainer'
fit(
  object,
  auto_calculate_gpu_ram = TRUE,
  force_refit = FALSE,
  test_one_fit = FALSE,
  ...
)

Arguments

object

A BatchBirdFlowTrainer().

auto_calculate_gpu_ram

Logical; if TRUE (default), sets gpu_ram := max(gpu_ram(trainer$bf), 8) before fitting.

force_refit

Logical; if TRUE, refit models even if identical HDF5s already exist.

test_one_fit

Logical; if TRUE, just do some test on this local session instead of submitting it to slurm

...

Additional arguments forwarded to the model fitting pipeline.

Value

The input trainer (invisible). Side effects: submits cluster jobs, writes HDF5 model files, and prunes extra preexisting fits when appropriate.

See also

birdflow_modelfit_args_df(), batch_modelfit_wrapper(), birdflow_modelfit()

Examples

if (FALSE) { # \dontrun{
trainer <- BatchBirdFlowTrainer("amewoo", res = 150)
fit(trainer, force_refit = FALSE)
} # }