WildTrain CLI Reference

Complete command-line interface reference for the wildtrain CLI.

WildTrain uses Typer with nested subcommand groups for training, evaluation, model registration, pipelines, and visualization.

wildtrain [COMMAND_GROUP] [COMMAND] [OPTIONS]

Global Options

Option	Description
`-v`, `--verbose`	Enable verbose logging
`-c`, `--config-dir`	Configuration directory
`--help`	Show help message and exit

`train` — Training Commands

Train detection and classification models.

`train classifier`

Train a classification model using PyTorch Lightning.

wildtrain train classifier [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to training configuration YAML file

Example:

wildtrain train classifier -c configs/classification/classification_train.yaml

See Classification Training Config for config details.

`train detector`

Train an object detection model (YOLO via Ultralytics).

wildtrain train detector [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to training configuration YAML file

Example:

wildtrain train detector -c configs/detection/yolo_configs/yolo_train.yaml

See Detection Training Config for config details.

`evaluate` — Evaluation Commands

Evaluate trained models on test/validation datasets.

`evaluate classifier`

Evaluate a classification model using a YAML config file.

wildtrain evaluate classifier [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to classification evaluation YAML config
`--debug`	`bool`	`false`	Enable debug mode

Example:

wildtrain evaluate classifier -c configs/classification/classification_eval.yaml

`evaluate detector`

Evaluate a YOLO detection model using a YAML config file.

wildtrain evaluate detector [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to YOLO evaluation YAML config
`--debug`	`bool`	`false`	Enable debug mode

Example:

wildtrain evaluate detector -c configs/detection/detection_sweep.yaml

`evaluate yolo-model`

Run direct Ultralytics YOLO model validation (bypass WildTrain wrapper).

wildtrain evaluate yolo-model [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to YOLO evaluation YAML config

The config YAML should contain a model key (path to weights) plus any Ultralytics val() parameters.

`register` — Model Registration Commands

`register classifier`

wildtrain register classifier [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`None`	Path to registration configuration file
`--weights-path`	`PATH`	`None`	Path to model checkpoint file
`-n`, `--name`	`str`	`classifier`	Model name for registration
`-b`, `--batch-size`	`int`	`8`	Batch size for inference
`--mlflow-uri`	`str`	`http://localhost:5000`	MLflow tracking server URI

You can use either --config or provide options directly. When using --config, don't provide other options.

Examples:

# Using config file
wildtrain register classifier -c configs/registration/classifier_registration_example.yaml

# Using direct options
wildtrain register classifier --weights-path model.ckpt --name my_classifier --mlflow-uri http://localhost:5000

See Registration Config for config details.

`register detector`

wildtrain register detector CONFIG_PATH

Argument	Type	Description
`CONFIG_PATH`	`PATH`	Path to detector registration configuration file

Example:

wildtrain register detector configs/registration/detector_registration_example.yaml

`pipeline` — Pipeline Commands

Run full train + eval pipelines in a single command.

`pipeline detection`

Run the full detection pipeline (train + evaluate).

wildtrain pipeline detection [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`None`	Path to unified detection pipeline YAML config

Example:

wildtrain pipeline detection -c configs/detection/detection_sweep.yaml

`pipeline classification`

Run the full classification pipeline (train + evaluate).

wildtrain pipeline classification [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`None`	Path to unified classification pipeline YAML config

Example:

wildtrain pipeline classification -c configs/classification/classification_pipeline_config.yaml

`config` — Configuration Management

Validate and generate configuration templates.

`config validate`

Validate a configuration file against Pydantic models.

wildtrain config validate CONFIG_PATH [OPTIONS]

Argument / Option	Type	Default	Description
`CONFIG_PATH`	`PATH`	(required)	Path to configuration file
`-t`, `--type`	`str`	`classification`	Config type (see below)

Supported config types: classification, detection, classification_eval, detection_eval, classification_visualization, detection_visualization, pipeline, detector_registration, classifier_registration, model_registration

Example:

wildtrain config validate configs/classification/classification_train.yaml --type classification

`config template`

Generate a default YAML configuration template.

wildtrain config template CONFIG_TYPE [OPTIONS]

Argument / Option	Type	Default	Description
`CONFIG_TYPE`	`str`	(required)	Configuration type to generate template for
`-s`, `--save`	`PATH`	`None`	Save template to file (prints to stdout if omitted)

Example:

# Print template to console
wildtrain config template classification

# Save to file
wildtrain config template detection -s my_detection_config.yaml

`dataset` — Dataset Commands

Dataset analysis and statistics.

`dataset stats`

Compute dataset statistics (mean, standard deviation) for normalization.

wildtrain dataset stats DATA_DIR [OPTIONS]

Argument / Option	Type	Default	Description
`DATA_DIR`	`PATH`	(required)	Path to dataset directory
`--split`	`str`	`train`	Split to compute statistics for
`-o`, `--output`	`PATH`	`None`	Output file for statistics JSON

Example:

wildtrain dataset stats D:/data/roi_dataset --split train -o stats.json

`visualize` — Visualization Commands

Upload model predictions to FiftyOne for interactive visualization.

`visualize classifier-predictions`

Upload classifier predictions to a FiftyOne dataset.

wildtrain visualize classifier-predictions [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to classification visualization YAML config

Example:

wildtrain visualize classifier-predictions -c configs/classification/classification_visualization.yaml

`visualize detector-predictions`

Upload detector predictions to a FiftyOne dataset.

wildtrain visualize detector-predictions [OPTIONS]

Option	Type	Default	Description
`-c`, `--config`	`PATH`	`""`	Path to detection visualization YAML config

Example:

wildtrain visualize detector-predictions -c configs/detection/visualization.yaml

`run-server` — Inference Server

Start a LitServe-based inference server for model serving.

wildtrain run-server [OPTIONS]

Option	Type	Default	Description
`--port`	`int`	`4141`	Port to run the server on
`-w`	`int`	`1`	Number of workers per device
`-c`, `--config`	`PATH`	`None`	Path to inference config file

When a config file is provided, it sets MLflow environment variables and overrides port/workers.

Example:

# Using config file
wildtrain run-server -c configs/inference.yaml

# Using direct options
wildtrain run-server --port 4141 -w 2

Quick Reference

Command	Description
`wildtrain train classifier -c CONFIG`	Train a classifier
`wildtrain train detector -c CONFIG`	Train a YOLO detector
`wildtrain evaluate classifier -c CONFIG`	Evaluate a classifier
`wildtrain evaluate detector -c CONFIG`	Evaluate a detector
`wildtrain register classifier -c CONFIG`	Register classifier to MLflow
`wildtrain register detector CONFIG`	Register detector to MLflow
`wildtrain pipeline detection -c CONFIG`	Full detection pipeline
`wildtrain pipeline classification -c CONFIG`	Full classification pipeline
`wildtrain config validate CONFIG --type TYPE`	Validate a config file
`wildtrain config template TYPE`	Generate config template
`wildtrain dataset stats DATA_DIR`	Compute dataset stats
`wildtrain visualize classifier-predictions -c CONFIG`	Visualize classifier predictions
`wildtrain visualize detector-predictions -c CONFIG`	Visualize detector predictions
`wildtrain run-server -c CONFIG`	Start inference server

For configuration file details, see Configuration Reference.
For shell scripts, see WildTrain Scripts.

Global Options​

train — Training Commands​

train classifier​

train detector​

evaluate — Evaluation Commands​

evaluate classifier​

evaluate detector​

evaluate yolo-model​

register — Model Registration Commands​

register classifier​

register detector​

pipeline — Pipeline Commands​

pipeline detection​

pipeline classification​

config — Configuration Management​

config validate​

config template​

dataset — Dataset Commands​

dataset stats​

visualize — Visualization Commands​

visualize classifier-predictions​

visualize detector-predictions​

run-server — Inference Server​

Quick Reference​

Global Options

`train` — Training Commands

`train classifier`

`train detector`

`evaluate` — Evaluation Commands

`evaluate classifier`

`evaluate detector`

`evaluate yolo-model`

`register` — Model Registration Commands

`register classifier`

`register detector`

`pipeline` — Pipeline Commands

`pipeline detection`

`pipeline classification`

`config` — Configuration Management

`config validate`

`config template`

`dataset` — Dataset Commands

`dataset stats`

`visualize` — Visualization Commands

`visualize classifier-predictions`

`visualize detector-predictions`

`run-server` — Inference Server

Quick Reference