Machine Learning workflows

Training, evaluation, optimization, inference, and model lifecycle patterns. Each workflow below has its own page with a DAG, inputs, outputs, provenance, and execution considerations.

Workflow	Purpose
Training with MLflow	Train and register a model while preserving data, parameters, metrics, and artifacts.
Hyperparameter search	Explore independent configurations in parallel and select one with a reproducible objective.
LoRA fine-tuning	Produce lightweight adapters from a pinned base model and instruction dataset.
Batch inference	Partition a large input, infer concurrently, and merge traceable predictions.
Comparative model evaluation	Evaluate candidate models with the same immutable dataset and policy.
Model conversion and quantization	Create hardware-specific variants and validate compatibility and quality.
Computer vision training	Expose collection, labels, augmentation, training, and packaging as observable stages.
Anomaly detection	Train a detector and make threshold selection an explicit workflow decision.
Model ensemble	Run independent predictors and combine them through a versioned aggregation rule.
Model distillation	Generate teacher targets and train a smaller student with measured quality and latency.
Inference benchmark	Compare one frozen model and input set across CPU, GPU, and edge targets.
Infrastructure selection for training	Compare plans for local GPU, HPC, and cloud before executing the selected candidate.