Diagnostics

Use the NeuRepTrace doctor command to check whether an installed environment can run the core workflows and whether dataset configuration files are structurally valid before launching long decoding jobs.

Environment check

neureptrace doctor

The command reports the Python version, the installed NeuRepTrace version, and the availability of required runtime dependencies. Optional machine-learning extras such as XGBoost and PyTorch are reported as warnings when they are not installed. Dependency availability is checked by importing the corresponding Python modules, so broken compiled extensions or incompatible binary packages are reported as failed diagnostics instead of being hidden behind package metadata.

Use a quieter core-only check in continuous-integration jobs or lightweight environments:

neureptrace doctor --skip-optional

For automation, request JSON output:

neureptrace doctor --skip-optional --json

Dataset configuration check

Dataset YAML or JSON files can be validated without loading the full neural data:

neureptrace doctor \
  --dataset-config examples/configs/pymegdec_bushmeg.yml \
  --skip-optional

Add --check-dataset-files when the data directory is mounted and all referenced files should already exist:

neureptrace doctor \
  --dataset-config configs/bush_meg/stimulus_decoding.yml \
  --check-dataset-files

Requiring project-specific modules

Some benchmark workflows depend on optional project-level modules or adapters. Use --require-module to make such imports fail the diagnostic explicitly:

neureptrace doctor \
  --require-module xgboost \
  --require-module torch

The command exits with status code 1 when required checks fail. It exits with status code 0 when only optional dependency warnings are present, unless --fail-on-warnings is supplied.