adds retrained Reisa LoRA

reisa/README.md

@@ -1,23 +1,49 @@
-# Reisa  (Blue Archive)
-Came out okay, but not great, largely due to a lack of quality training data. Very overfit on her outfit so I recommend dropping the weight to 0.85 to dress her up in other stuff.
+# Uzawa Reisa (Blue Archive)
+宇沢レイサ (ブルーアーカイブ) / 우자와 레이사 (블루 아카이브)
+
+[**Download here.**](https://huggingface.co/khanon/lora-training/blob/main/reisa/chara-reisa-v3.safetensors)
+
+## Table of Contents
+- [Preview](#preview)
+- [Usage](#usage)
+- [Training](#training)
+- [Revisions](#revisions)
+
+## Preview
+
+![Reisa portrait](chara-reisa-v3.png)
+![Reisa V](example-001-AOM2NMMElyv3.png)
+![Reisa pose](example-002-DefmixRed-v3.png) 
+
+Unfortunately the model frequently struggles with Reisa's halo.
 
 ## Usage
-Use any or all of these tags to summon Reisa:
-`reisa, halo, low twintails, pink hair, purple eyes, star hair ornament`
-Hair and eyes are mostly optional if you describe a bit of her outfit as well.
-She naturally likes to make her `:D` expression because most art features her doing that.  However I also included images tagged with other expressions and you can get her to perform other expressions reasonably easily.
 
-For her normal outfit (she will probably show up in this by default, so negative prompt these to change her)
-`school uniform, grey shirt, blue jacket, pleated skirt, striped socks, backpack, sneakers`
+Use any or all of these tags to summon Reisa: `reisa, halo, low twintails, pink hair, purple eyes, star hair ornament`
+- Hair and eye tags are optional.
+
+She likes to make her `:d` expression by default.  I included images tagged with other expressions, so you should be able to change it, but if not try adding `:d` to the negative prompt.
 
-I suggest lowering the weight or using epoch 6 if she's really fighting against you changing her outfit.
+For her normal outfit: `school uniform, grey shirt, blue jacket, pleated skirt, striped socks, backpack, sneakers`
 
 ## Training
-*All parameters are provided in the accompanying JSON files.*
-- Trained on a set of 89 images, repeated 17 times 7 epochs (89 images * 17 repeats / 3 batch size * 7 epochs = 3530 steps)
-  - Dataset included a mixture of SFW and NSFW.
-- Initially tagged with WD1.4 Convnextv2 model, then performed heavy pruning and editing.
-  - Made sure important traits were present and consitently described, and traits like `halo` were consistent with actual visibility
-  - Added many facial expression, camera angle, and image composition hints
-- Used network_dimension 128 (same as usual) and network_alpha 64
+*Exact parameters are provided in the accompanying JSON files.*
+- Trained on a set of 89 images
+  - Datset included a mixture of SFW/NSFW
+  - 13 repeats
+  - 3 batch size, 4 epochs
+  - `(89*13) / 3 * 4` = 1543 steps
+- 768x768 training resolution
+- `constant_with_warmup` scheduler instead of `cosine`
+- Initially tagged with WD1.4 Convnextv2 model. Tags were edited for accuracy and improved outfit consistency.  Tags for facial expressions, camera angle, and image composition were added.
+- Used network_dimension 128 (same as usual) / network alpha 128 (default)
 - Trained without VAE.
+- Training dataset available here. 
+
+## Revisions
+- v3 (2023-02-06)
+  - Completely re-trained to correct issue where caption files were not being loaded correctly.  New version should have more flexibility but requires additional outfit prompting.
+- v2b (2023-01-28)
+  - Initial release.
+  - **Issue:** caption files were completely ignored during training. As a result, prompting "reisa" grants a very strong effect, but the LoRA is highly overfit.
+  - [Download available on Mega.co.nz repository.](https://mega.nz/folder/yjwGiIBD#w79CxW2-Cj66QmObg0zOdg)

reisa/chara-reisa-v3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ef0d52aeff5134ecf5755b6a3e59275a344d30a5bf3ef61982c2ef73099afe9
+size 151118356

reisa/lora_chara_reisa_v2_3530step_832_batch3_5e-5text_2e-4unet_7epoch.json

@@ -1,50 +0,0 @@
-{
-  "pretrained_model_name_or_path": "G:/sd/repo/models/Stable-diffusion/nai-animefull-final-pruned.safetensors",
-  "v2": false,
-  "v_parameterization": false,
-  "logging_dir": "",
-  "train_data_dir": "G:/sd/training/datasets/reisa/dataset",
-  "reg_data_dir": "G:/sd/training/datasets/regempty",
-  "output_dir": "G:/sd/lora/trained/chara/reisa",
-  "max_resolution": "832,832",
-  "learning_rate": "",
-  "lr_scheduler": "cosine_with_restarts",
-  "lr_warmup": "5",
-  "train_batch_size": 3,
-  "epoch": "7",
-  "save_every_n_epochs": "2",
-  "mixed_precision": "fp16",
-  "save_precision": "fp16",
-  "seed": "31337",
-  "num_cpu_threads_per_process": 32,
-  "cache_latents": true,
-  "caption_extension": "",
-  "enable_bucket": true,
-  "gradient_checkpointing": false,
-  "full_fp16": false,
-  "no_token_padding": false,
-  "stop_text_encoder_training": 0,
-  "use_8bit_adam": true,
-  "xformers": true,
-  "save_model_as": "safetensors",
-  "shuffle_caption": true,
-  "save_state": false,
-  "resume": "",
-  "prior_loss_weight": 1.0,
-  "text_encoder_lr": "5e-5",
-  "unet_lr": "2e-4",
-  "network_dim": 128,
-  "lora_network_weights": "",
-  "color_aug": false,
-  "flip_aug": false,
-  "clip_skip": 2,
-  "gradient_accumulation_steps": 1.0,
-  "mem_eff_attn": false,
-  "output_name": "chara-reisa-v2",
-  "model_list": "",
-  "max_token_length": "150",
-  "max_train_epochs": "",
-  "max_data_loader_n_workers": "",
-  "network_alpha": 64,
-  "training_comment": ""
-}

reisa/{lora_chara_reisa_v1_3115step_768_batch3_5e-5text_2e-4unet_7epoch.json → lora_chara_reisa_v3_89i13r.json}

@@ -4,21 +4,21 @@
   "v_parameterization": false,
   "logging_dir": "",
   "train_data_dir": "G:/sd/training/datasets/reisa/dataset",
-  "reg_data_dir": "G:/sd/training/datasets/regempty",
+  "reg_data_dir": "",
   "output_dir": "G:/sd/lora/trained/chara/reisa",
   "max_resolution": "768,768",
-  "learning_rate": "",
-  "lr_scheduler": "cosine_with_restarts",
+  "learning_rate": "1e-5",
+  "lr_scheduler": "constant_with_warmup",
   "lr_warmup": "5",
   "train_batch_size": 3,
-  "epoch": "7",
-  "save_every_n_epochs": "2",
+  "epoch": "4",
+  "save_every_n_epochs": "",
   "mixed_precision": "fp16",
   "save_precision": "fp16",
   "seed": "31337",
   "num_cpu_threads_per_process": 32,
   "cache_latents": true,
-  "caption_extension": "",
+  "caption_extension": ".txt",
   "enable_bucket": true,
   "gradient_checkpointing": false,
   "full_fp16": false,
@@ -31,8 +31,8 @@
   "save_state": false,
   "resume": "",
   "prior_loss_weight": 1.0,
-  "text_encoder_lr": "5e-5",
-  "unet_lr": "2e-4",
+  "text_encoder_lr": "1.5e-5",
+  "unet_lr": "1.5e-4",
   "network_dim": 128,
   "lora_network_weights": "",
   "color_aug": false,
@@ -40,11 +40,15 @@
   "clip_skip": 2,
   "gradient_accumulation_steps": 1.0,
   "mem_eff_attn": false,
-  "output_name": "chara-reisa-v1",
+  "output_name": "chara-reisa-v3",
   "model_list": "",
   "max_token_length": "150",
   "max_train_epochs": "",
   "max_data_loader_n_workers": "",
-  "network_alpha": 64,
-  "training_comment": ""
+  "network_alpha": 128,
+  "training_comment": "Character: `reisa, halo, low twintails, pink hair, purple eyes, star hair ornament` (hair and eyes optional)\nOutfit: `school uniform, grey shirt, blue jacket, pleated skirt, striped socks, backpack, sneakers`\n\n89 images * 13 repeats / 3 batch size * 4 epochs = 1543 steps",
+  "keep_tokens": 2,
+  "lr_scheduler_num_cycles": "",
+  "lr_scheduler_power": "",
+  "persistent_data_loader_workers": true
 }