Update d-adaptation/notes.md

d-adaptation/notes.md

@@ -22,7 +22,7 @@ Lower dims show good performance. Need much larger test to check for accuracy be
 ## Resolution
 512 resolution is good so far 700 steps.
 640 resolution is good. 900 steps. UMP has 640 resolution sample.
-768 resolution at 1800 steps was not good at all. After 800 steps in it melted and became a mess but it was also when it started to pick up the features of the character.
+768 resolution has not been doing well with the 1.0/0.5 LR. It melts. Given 640 was also melting with 1.0/0.5 and higher cycles, it is likely leraning rate neesd to be scaled down. There may still be a value that works for all three, but it is sensitive to starting LRs.
 
 ## Steps
 600 to 800 steps have been the primary focus of experimentation.