From d853c5b356f352ab71056d376e48580d7bd3225d Mon Sep 17 00:00:00 2001
From: EricLBuehler <ericlbuehler@gmail.com>
Date: Mon, 11 Sep 2023 15:32:10 -0400
Subject: [PATCH] Update readme

---
 README.md | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 0c9a16e..13ed2a9 100644
--- a/README.md
+++ b/README.md
@@ -14,7 +14,14 @@ candle-lora is able to convert:
 
 All conversions are done as implemented in HuggingFace's official LoRA implementation.
 
-Current working example:
+## How to use
+1) Replace any concrete `Linear`, `Conv1d`, `Conv2d`, or `Embedding` types with `Box<dyn ...LayerLike>`. This will allow `candle-lora` to
+generate new layers that can easily be swaped out without forcing you to redefine your model structs.
+2) Select the layers and perform the conversion.
+3) Swap out the layers.
+4) Enjoy your new LoRA model!
+
+## Example
 ```rust
 use std::{collections::HashMap, hash::Hash};