Training
Supervised fine-tuning (SFT)
Edit or adapt a modelFine-tuningReferenceInstruction / domain tuning: keep training the model on curated input→output pairs to teach a task or style.
- Objective
- Edit or adapt a model
- Targets
- Weights
- Format
- Next-token training on curated data
- Granularity
- Full-weight or adapter
- Lifecycle
- QAT / QAD · needs gradients
- Calibration
- Full corpus
- Compression
- None — reshapes behavior, not size
- Quality
- Specializes behavior; risks catastrophic forgetting if over-tuned
- Hardware
- Training run — GPU-hours; full-weight needs the model in memory twice
- Runtimes
A worked Spark recipe for this method hasn't been written yet — it lives here as a reference point in the ontology.