From 31a0f3944ce7375dd968ab585cb84ebbe0f5e7f0 Mon Sep 17 00:00:00 2001
From: PENG Bo <33809201+BlinkDL@users.noreply.github.com>
Date: Wed, 13 Apr 2022 15:58:35 +0800
Subject: [PATCH] Update README.md

---
 README.md | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/README.md b/README.md
index fd4a1fd..53343fa 100644
--- a/README.md
+++ b/README.md
@@ -36,6 +36,8 @@ I propose a simple new method to find better LR schedules. The method is cost-ef
 
 UPDATE: In "Conclusion 1.", remember to use the best-fitting regime (ignore the initial steps where our approximations break down) to fit the parameters.
 
+In the last three plots, black = predicted loss curve of the new LR schedule, blue = original (unoptimized) real loss curve, orange = new LR schedule.
+
 # The top-p-x sampling method
 
 We propose a new sampling method called top-p-x: