Update README.md
Browse files
README.md
CHANGED
|
@@ -81,4 +81,5 @@ As we compare the outputs of the models directly, and not to an absolute expecte
|
|
| 81 |
|
| 82 |
The merge uses a value for lambda that gives equal weight to both models. The gradients of Animagine are on average larger than those of NoobAI, so we need to tweak lambda if we want a merge that includes a proportional amount of parameters from both models. I am not aware of a closed form solution for lambda, so I used a bisecting algorithm to estimate it to a reasonable precision.
|
| 83 |
|
| 84 |
-
This gave a value of lambda = 0.224609375, which corresponds to the "DeadCenter" version.
|
|
|
|
|
|
| 81 |
|
| 82 |
The merge uses a value for lambda that gives equal weight to both models. The gradients of Animagine are on average larger than those of NoobAI, so we need to tweak lambda if we want a merge that includes a proportional amount of parameters from both models. I am not aware of a closed form solution for lambda, so I used a bisecting algorithm to estimate it to a reasonable precision.
|
| 83 |
|
| 84 |
+
- This gave a value of lambda = 0.224609375, which corresponds to the "DeadCenter" version.
|
| 85 |
+
- Another opiniated value of lambda that I found aesthetically pleasing with some artist tags is lambda = 0.3564453125 (more a random value than an optimized one), which corresponds to the "MadHatter" version.
|