Merge branch 'main' of https://huggingface.co/alea31415/LyCORIS-experiments into main
Browse files
README.md
CHANGED
|
@@ -37,12 +37,27 @@ For a thorough comparaison please refer to the `generated_samples` folder.
|
|
| 37 |
Dataset, in general, is the most important out of all.
|
| 38 |
The common wisdom that we should prune anything that we want to be attach to the trigger word is exactly the way to go for.
|
| 39 |
No tags at all (top three rows) is terrible, especially for style training.
|
| 40 |
-
Having all the tags (bottom three rows) remove the traits from subjects if these tags are not used during sampling (not completely true but more or less the case).
|
| 41 |
|
| 42 |
|
| 43 |
|
| 44 |
|
| 45 |
-
####
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
The most prominent benefit of training at higher resolution is that it helps generating more complex/detailed background.
|
| 48 |
Chances are that you can get more details about the outfit or pupils etc.
|
|
|
|
| 37 |
Dataset, in general, is the most important out of all.
|
| 38 |
The common wisdom that we should prune anything that we want to be attach to the trigger word is exactly the way to go for.
|
| 39 |
No tags at all (top three rows) is terrible, especially for style training.
|
| 40 |
+
Having all the tags (bottom three rows) remove the traits from subjects if these tags are not used during sampling (not completely true but more or less the case, see also discussion below).
|
| 41 |
|
| 42 |
|
| 43 |
|
| 44 |
|
| 45 |
+
#### The effect of style images on characters
|
| 46 |
+
|
| 47 |
+
I do beleive regularization images are important, far more important than tweaking any hyperparameters. They slow down training but also make sure that the undesired aspect are less baked into the model if we have images of other types, even if they are not for the subjects we train for.
|
| 48 |
+
|
| 49 |
+
Comparing the models trained with and without style images, we can see that models trained with general style images have less anime styles baked in. The difference is particularly clear for Tilty, who only have anime screenshots for training.
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
On the other hand, the default clothes seem to be better trained when there is no regularization image. While this may seem beneficial, it is worth noticing that I keep all the output tags. Therefore, in a sense we only want to get the outputs when we prompt them explicitly. The magic of having the trigger words to fill in what is not in caption seems to be more pronouncing when we have regularization images. In any case, this magic will not work forever as we will eventually start overfitting. The following image show that we get images that are much closer after putting clothes in prompts.
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
In any case, if your regularization images are properly tagged with of a lot of concepts, then you always have the benefit that you can combine them with the main things you train for.
|
| 58 |
+
|
| 59 |
+
|
| 60 |
+
#### Training resolution
|
| 61 |
|
| 62 |
The most prominent benefit of training at higher resolution is that it helps generating more complex/detailed background.
|
| 63 |
Chances are that you can get more details about the outfit or pupils etc.
|