Language-model training and adaptation · MLD-0968
QLoRA
Fine-tuning LoRA adapters while a frozen base model is held in a low-bit representation, commonly using 4-bit quantization.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Adapt a larger model within a smaller memory budget.
Original caveat
Quantized storage does not imply all arithmetic uses four-bit precision.
Context and aliases
Level: Intermediate. Stages: Training · Adaptation · Evaluation.
Tags: DL · SLM · LLM · NLP.
Aliases: Quantized low-rank adaptation.
Related concepts
Concepts linking here
Original references
- PEFT: LoRA conceptual guide ↗ · Consulted
- PEFT: quantization guide ↗ · Consulted
- Original reference URL ↗