AdvancedVocabulary#ai-llm#data-science-ml#developer-tools

Low-Rank Adaptation (LoRA) Vocabulary

Build fluency in the vocabulary of fine-tuning a model through small, injected low-rank matrices.

0 / 5 completed
1 / 5
At standup, a dev mentions fine-tuning a large pretrained model by injecting a pair of small trainable low-rank matrices alongside its frozen original weights, instead of updating every one of the model's parameters. What is this technique called?