quuMerge branch 'main' of https://huggingface.co/alea31415/LyCORIS-experiments into main

.gitattributes

@@ -110,3 +110,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 0325_aniscreen_fanart_styles/0325exp-cs1-tte1e-4-step-12000.safetensors filter=lfs diff=lfs merge=lfs -text
 0325_aniscreen_fanart_styles/0325exp-cs2-step-6000.safetensors filter=lfs diff=lfs merge=lfs -text
 0324_all_aniscreen_tags/0324exp-cs1.safetensors filter=lfs diff=lfs merge=lfs -text
+generated_samples/00165-20230327023658.jpg filter=lfs diff=lfs merge=lfs -text
+generated_samples/00165-20230327023658.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -2,7 +2,7 @@
 license: creativeml-openrail-m
 ---
 
-Tags to test
+### Trigger words
 
 ```
 Anisphia, Euphyllia, Tilty, OyamaMahiro, OyamaMihari
@@ -14,7 +14,7 @@ For `0324_all_aniscreen_tags`, I accidentally tag all the character images with
 For `0325_aniscreen_fanart_styles`, things are done correctly (anime screenshots tagged as `aniscreen`, fanart tagged as `fanart`).
 
 
-#### Settings
+### Settings
 
 Default settings are
 - loha net dim 8, conv dim 4, alpha 1
@@ -28,7 +28,28 @@ The configuration json files can otherwsie be found in the `config` subdirectori
 However, some experiments concern the effect of tags for which I regenerate the txt file and the difference can not be seen from the configuration file in this case.
 For now this concerns `05tag` for which tags are only used with probability 0.5.
 
-#### Datasets
+### Some observations
+
+For a thorough comparaison please refer to the `generated_samples` folder.
+
+#### Captioning
+
+Dataset, in general, is the most important out of all.
+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.
+No tags at all is terrible, especially for style training.
+Having all the tags remove the traits from subjects if these tags are not used during sampling (not completely true but more or less the case).
+
+![00066-20230326090858](https://huggingface.co/alea31415/LyCORIS-experiments/resolve/main/generated_samples/00066-20230326090858.png)
+
+
+#### Others
+
+1. I barely see any difference for training at clip skip 1 and 2.
+2. Setting text encoder learning rate to be half of that of unet makes training two times slower while I cannot see how it helps.
+3. The difference between lora, locon, and loha are very subtle.
+4. Training at higher resolution helps generating more complex backgrounds etc, but it is very time-consuming and most of the time it isn't worth it (simpler to just switch base model) unless this is exactly the goal of the lora you're training.
+
+### Datasets
 
 Here is the composition of the datasets
 ```
@@ -51,6 +72,4 @@ Here is the composition of the datasets
 5_artists~onono_imoko: 373
 7_characters~screenshots~Tilty: 95
 9_characters~fanart~Anisphia+Euphyllia: 97
-```
-
-Tks
+```

generated_samples/00165-20230327023658.txt

@@ -0,0 +1,6 @@
+OyamaMahiro; fanart; (w-sitting, wariza:1.1), 1girl, solo,   skirt,  eating lollipop 
+ looking at viewer, detailed background, outdoors, in village, cloud
+<lora:0325exp-cs1:1>
+Negative prompt: aniscreen, far from viewer, realistic, bad hands, lowres, bad anatomy,  text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry
+Steps: 25, Sampler: Euler a, CFG scale: 8, Seed: 1234, Size: 520x680, Model hash: 89d59c3dde, Model: base_nai-full-pruned, Script: X/Y/Z plot, X Type: Prompt S/R, X Values: "<lora:0325exp-cs1:1>,<lora:0325exp-cs1-tte1e-4:1>,<lora:0326exp-cs1-lora:1>,
+<lora:0325exp-cs1-768-step-28000:1>, <lora:0326exp-cs1-5e-4:1>,<lora:0326exp-cs1-dim32-16:1>, <lora:0326exp-cs1-dim32-16-alpha-half:1>, <lora:0326exp-cs1-dim32-16-5e-4:1>, <lora:0326exp-cs1-character-only:1>", Y Type: Seed, Y Values: "1234,555,98,444,343242", Fixed Y Values: "1234, 555, 98, 444, 343242"