nipype — Nipype execution settings

Schema: NipypeConfig in src/thesis/core/config/validators.py. Drives NipypeExecutor and the global nipype.config overrides.

Field	Type	Default	Description
working_dir	Path	work	Base working directory. Each patient workflow gets a <working_dir>/<patient_id>/ subdirectory. Processed through to_path() (supports ~, $VAR).
crash_dir	Path | None	None	Optional crash-dump directory. When unset, Nipype defaults apply.
plugin	str	"MultiProc"	Nipype execution plugin (MultiProc, Linear, SGE, PBS, …).
plugin_args	Dict[str, Any]	{}	Plugin arguments passed to workflow.run(). Common keys: n_procs, memory_gb, n_gpu_procs, n_gpus. The CLI auto-injects hardware.* and --max-workers values when not already set.
stop_on_first_crash	bool	False	Stop execution after the first crashed node.
remove_unnecessary_outputs	bool	True	Have Nipype clean intermediate outputs after workflow completion.
keep_inputs	bool	True	Keep input files in the node working directories.
hash_method	str	"content"	Caching hash strategy. content re-hashes file contents (slower, robust); timestamp uses mtime (faster, fragile). content is required for reliable cache reuse across batch retries.
use_profiler	bool	False	Enable Nipype’s resource profiler.

Example

nipype:
  working_dir: work
  crash_dir: work/crashes
  plugin: MultiProc
  plugin_args:
    n_procs: 8
    memory_gb: 32
  stop_on_first_crash: false
  remove_unnecessary_outputs: true
  keep_inputs: true
  hash_method: content
  use_profiler: false

Notes

• Both per-workflow execution (via NipypeExecutor) and the batch meta-workflow path (in cli.py) call apply_nipype_execution_config(), which writes the relevant settings into the global nipype.config. Without that, file-input hashing falls back to Nipype’s defaults — keep hash_method: content and let the framework apply the global update.

• The batch path also calls build_nipype_status_callback() to translate node start/finish/failure events into the structured CLI progress UI.

• When --max-workers N (or -j N) is passed, the CLI sets plugin_args["n_procs"] = N, overriding any YAML value.

• When hardware.gpu_enabled: true, the CLI sets plugin_args.setdefault("n_gpu_procs", hardware.n_gpu_procs) and plugin_args.setdefault("n_gpus", hardware.n_gpus).