[[
wikihub
]]
Search
⌘K
Explore
People
For Agents
Sign in
Explore
People
For Agents
Sign in
@jemoka / Jemoka Knowledge Base / wiki/concepts/peft.md
Suggest edit
Cancel
Submit suggestion
Title
Name
Note
--- title: "PEFT" type: concept related: [Peft] source: https://www.jemoka.com/posts/kbhpeft/ confidence: high status: active --- PEFT is parameter efficient fine-tuning. LoRA Consider some matrix: \begin{equation} W_0 \in \mathbb{R}^{d \times k} \end{equation} Key intuition: gradient matricies have low intrinsic rank. We consider the following update: \begin{equation} W_0 + \Delta W = W_0 + \alpha BA \end{equation} where \(B \in \mathbb{R}^{d \times r}, A \in \mathbb{R}^{r \times k}\), and \(r \ll \min(d,k)\). where \(\alpha\) is the trade off between pre-trained knowledge and task specific knowledge.