Language-model training and adaptation · MLD-0988
Rank-stabilized LoRA
A LoRA variant using a scaling rule designed to remain effective at higher adapter ranks.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Train higher-rank adapters more stably.
Original caveat
It changes scaling, not the low-rank update structure itself.
Context and aliases
Level: Advanced. Stages: Training · Adaptation · Evaluation.
Tags: DL · SLM · LLM · NLP.
Aliases: rsLoRA.
Related concepts
Original references
- Transformers glossary ↗ · Consulted
- PEFT: LoRA conceptual guide ↗ · Consulted
- TRL: supervised fine-tuning ↗ · Reference link
- TRL documentation ↗ · Consulted
- Original reference URL ↗