From 577373c4a8c90b3c9544b820ef66b176b8846323 Mon Sep 17 00:00:00 2001
From: AUTOMATIC1111 <16777216c@gmail.com>
Date: Tue, 11 Oct 2022 16:38:39 +0300
Subject: [PATCH] Updated Textual Inversion (markdown)

---
 Textual-Inversion.md | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/Textual-Inversion.md b/Textual-Inversion.md
index f904c7a..e87320d 100644
--- a/Textual-Inversion.md
+++ b/Textual-Inversion.md
@@ -86,3 +86,13 @@ Other options are to train on colabs and/or using diffusers library, which I kno
 - [viper1](https://gitgud.io/viper1/stable-diffusion-embeddings) - NSFW, furry girls.
 - [anon's embeddings](https://mega.nz/folder/7k0R2arB#5_u6PYfdn-ZS7sRdoecD2A) - NSFW, anime artists.
 - [rentry](https://rentry.org/embeddings) - a page with links to embeddings from many sources.
+
+# Hypernetworks
+
+Hypernetworks is a novel (get it?) concept for fine tuning a model without touching any of its weights.
+
+The current way to train hypernets is in the textual inversion tab.
+
+Training works the same way as with textual inversion.
+
+The only requirement is to use a very, very low learning rate, something like 0.000005 or 0.0000005.