Skip to content

Commit 7ffc83a

Browse files
Merge pull request #3 from kohya-ss/dev
updates
2 parents 4511da6 + 0d96e10 commit 7ffc83a

27 files changed

+1308
-317
lines changed

.github/workflows/typos.yml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -18,4 +18,4 @@ jobs:
1818
- uses: actions/checkout@v4
1919

2020
- name: typos-action
21-
uses: crate-ci/typos@v1.17.2
21+
uses: crate-ci/typos@v1.19.0

README.md

Lines changed: 130 additions & 7 deletions
Large diffs are not rendered by default.

docs/masked_loss_README-ja.md

Lines changed: 57 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,57 @@
1+
## マスクロスについて
2+
3+
マスクロスは、入力画像のマスクで指定された部分だけ損失計算することで、画像の一部分だけを学習することができる機能です。
4+
たとえばキャラクタを学習したい場合、キャラクタ部分だけをマスクして学習することで、背景を無視して学習することができます。
5+
6+
マスクロスのマスクには、二種類の指定方法があります。
7+
8+
- マスク画像を用いる方法
9+
- 透明度(アルファチャネル)を使用する方法
10+
11+
なお、サンプルは [ずんずんPJイラスト/3Dデータ](https://zunko.jp/con_illust.html) の「AI画像モデル用学習データ」を使用しています。
12+
13+
### マスク画像を用いる方法
14+
15+
学習画像それぞれに対応するマスク画像を用意する方法です。学習画像と同じファイル名のマスク画像を用意し、それを学習画像と別のディレクトリに保存します。
16+
17+
- 学習画像
18+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/607c5116-5f62-47de-8b66-9c4a597f0441)
19+
- マスク画像
20+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/53e9b0f8-a4bf-49ed-882d-4026f84e8450)
21+
22+
```.toml
23+
[[datasets.subsets]]
24+
image_dir = "/path/to/a_zundamon"
25+
caption_extension = ".txt"
26+
conditioning_data_dir = "/path/to/a_zundamon_mask"
27+
num_repeats = 8
28+
```
29+
30+
マスク画像は、学習画像と同じサイズで、学習する部分を白、無視する部分を黒で描画します。グレースケールにも対応しています(127 ならロス重みが 0.5 になります)。なお、正確にはマスク画像の R チャネルが用いられます。
31+
32+
DreamBooth 方式の dataset で、`conditioning_data_dir` で指定したディレクトリにマスク画像を保存してください。ControlNet のデータセットと同じですので、詳細は [ControlNet-LLLite](train_lllite_README-ja.md#データセットの準備) を参照してください。
33+
34+
### 透明度(アルファチャネル)を使用する方法
35+
36+
学習画像の透明度(アルファチャネル)がマスクとして使用されます。透明度が 0 の部分は無視され、255 の部分は学習されます。半透明の場合は、その透明度に応じてロス重みが変化します(127 ならおおむね 0.5)。
37+
38+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/0baa129b-446a-4aac-b98c-7208efb0e75e)
39+
40+
※それぞれの画像は透過PNG
41+
42+
学習時のスクリプトのオプションに `--alpha_mask` を指定するか、dataset の設定ファイルの subset で、`alpha_mask` を指定してください。たとえば、以下のようになります。
43+
44+
```toml
45+
[[datasets.subsets]]
46+
image_dir = "/path/to/image/dir"
47+
caption_extension = ".txt"
48+
num_repeats = 8
49+
alpha_mask = true
50+
```
51+
52+
## 学習時の注意事項
53+
54+
- 現時点では DreamBooth 方式の dataset のみ対応しています。
55+
- マスクは latents のサイズ、つまり 1/8 に縮小されてから適用されます。そのため、細かい部分(たとえばアホ毛やイヤリングなど)はうまく学習できない可能性があります。マスクをわずかに拡張するなどの工夫が必要かもしれません。
56+
- マスクロスを用いる場合、学習対象外の部分をキャプションに含める必要はないかもしれません。(要検証)
57+
- `alpha_mask` の場合、マスクの有無を切り替えると latents キャッシュが自動的に再生成されます。

docs/masked_loss_README.md

Lines changed: 56 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,56 @@
1+
## Masked Loss
2+
3+
Masked loss is a feature that allows you to train only part of an image by calculating the loss only for the part specified by the mask of the input image. For example, if you want to train a character, you can train only the character part by masking it, ignoring the background.
4+
5+
There are two ways to specify the mask for masked loss.
6+
7+
- Using a mask image
8+
- Using transparency (alpha channel) of the image
9+
10+
The sample uses the "AI image model training data" from [ZunZunPJ Illustration/3D Data](https://zunko.jp/con_illust.html).
11+
12+
### Using a mask image
13+
14+
This is a method of preparing a mask image corresponding to each training image. Prepare a mask image with the same file name as the training image and save it in a different directory from the training image.
15+
16+
- Training image
17+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/607c5116-5f62-47de-8b66-9c4a597f0441)
18+
- Mask image
19+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/53e9b0f8-a4bf-49ed-882d-4026f84e8450)
20+
21+
```.toml
22+
[[datasets.subsets]]
23+
image_dir = "/path/to/a_zundamon"
24+
caption_extension = ".txt"
25+
conditioning_data_dir = "/path/to/a_zundamon_mask"
26+
num_repeats = 8
27+
```
28+
29+
The mask image is the same size as the training image, with the part to be trained drawn in white and the part to be ignored in black. It also supports grayscale (127 gives a loss weight of 0.5). The R channel of the mask image is used currently.
30+
31+
Use the dataset in the DreamBooth method, and save the mask image in the directory specified by `conditioning_data_dir`. It is the same as the ControlNet dataset, so please refer to [ControlNet-LLLite](train_lllite_README.md#Preparing-the-dataset) for details.
32+
33+
### Using transparency (alpha channel) of the image
34+
35+
The transparency (alpha channel) of the training image is used as a mask. The part with transparency 0 is ignored, the part with transparency 255 is trained. For semi-transparent parts, the loss weight changes according to the transparency (127 gives a weight of about 0.5).
36+
37+
![image](https://github.com/kohya-ss/sd-scripts/assets/52813779/0baa129b-446a-4aac-b98c-7208efb0e75e)
38+
39+
※Each image is a transparent PNG
40+
41+
Specify `--alpha_mask` in the training script options or specify `alpha_mask` in the subset of the dataset configuration file. For example, it will look like this.
42+
43+
```toml
44+
[[datasets.subsets]]
45+
image_dir = "/path/to/image/dir"
46+
caption_extension = ".txt"
47+
num_repeats = 8
48+
alpha_mask = true
49+
```
50+
51+
## Notes on training
52+
53+
- At the moment, only the dataset in the DreamBooth method is supported.
54+
- The mask is applied after the size is reduced to 1/8, which is the size of the latents. Therefore, fine details (such as ahoge or earrings) may not be learned well. Some dilations of the mask may be necessary.
55+
- If using masked loss, it may not be necessary to include parts that are not to be trained in the caption. (To be verified)
56+
- In the case of `alpha_mask`, the latents cache is automatically regenerated when the enable/disable state of the mask is switched.

docs/train_network_README-ja.md

Lines changed: 9 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -102,6 +102,8 @@ accelerate launch --num_cpu_threads_per_process 1 train_network.py
102102
* Text Encoderに関連するLoRAモジュールに、通常の学習率(--learning_rateオプションで指定)とは異なる学習率を使う時に指定します。Text Encoderのほうを若干低めの学習率(5e-5など)にしたほうが良い、という話もあるようです。
103103
* `--network_args`
104104
* 複数の引数を指定できます。後述します。
105+
* `--alpha_mask`
106+
* 画像のアルファ値をマスクとして使用します。透過画像を学習する際に使用します。[PR #1223](https://github.com/kohya-ss/sd-scripts/pull/1223)
105107

106108
`--network_train_unet_only``--network_train_text_encoder_only` の両方とも未指定時(デフォルト)はText EncoderとU-Netの両方のLoRAモジュールを有効にします。
107109

@@ -181,16 +183,16 @@ python networks\extract_lora_from_dylora.py --model "foldername/dylora-model.saf
181183

182184
詳細は[PR #355](https://github.com/kohya-ss/sd-scripts/pull/355) をご覧ください。
183185

184-
SDXLは現在サポートしていません。
185-
186186
フルモデルの25個のブロックの重みを指定できます。最初のブロックに該当するLoRAは存在しませんが、階層別LoRA適用等との互換性のために25個としています。またconv2d3x3に拡張しない場合も一部のブロックにはLoRAが存在しませんが、記述を統一するため常に25個の値を指定してください。
187187

188+
SDXL では down/up 9 個、middle 3 個の値を指定してください。
189+
188190
`--network_args` で以下の引数を指定してください。
189191

190192
- `down_lr_weight` : U-Netのdown blocksの学習率の重みを指定します。以下が指定可能です。
191-
- ブロックごとの重み : `"down_lr_weight=0,0,0,0,0,0,1,1,1,1,1,1"` のように12個の数値を指定します
193+
- ブロックごとの重み : `"down_lr_weight=0,0,0,0,0,0,1,1,1,1,1,1"` のように12個(SDXL では 9 個)の数値を指定します
192194
- プリセットからの指定 : `"down_lr_weight=sine"` のように指定します(サインカーブで重みを指定します)。sine, cosine, linear, reverse_linear, zeros が指定可能です。また `"down_lr_weight=cosine+.25"` のように `+数値` を追加すると、指定した数値を加算します(0.25~1.25になります)。
193-
- `mid_lr_weight` : U-Netのmid blockの学習率の重みを指定します。`"down_lr_weight=0.5"` のように数値を一つだけ指定します。
195+
- `mid_lr_weight` : U-Netのmid blockの学習率の重みを指定します。`"down_lr_weight=0.5"` のように数値を一つだけ指定します(SDXL の場合は 3 個)
194196
- `up_lr_weight` : U-Netのup blocksの学習率の重みを指定します。down_lr_weightと同様です。
195197
- 指定を省略した部分は1.0として扱われます。また重みを0にするとそのブロックのLoRAモジュールは作成されません。
196198
- `block_lr_zero_threshold` : 重みがこの値以下の場合、LoRAモジュールを作成しません。デフォルトは0です。
@@ -215,6 +217,9 @@ network_args = [ "block_lr_zero_threshold=0.1", "down_lr_weight=sine+.5", "mid_l
215217

216218
フルモデルの25個のブロックのdim (rank)を指定できます。階層別学習率と同様に一部のブロックにはLoRAが存在しない場合がありますが、常に25個の値を指定してください。
217219

220+
SDXL では 23 個の値を指定してください。一部のブロックにはLoRA が存在しませんが、`sdxl_train.py`[階層別学習率](./train_SDXL-en.md) との互換性のためです。
221+
対応は、`0: time/label embed, 1-9: input blocks 0-8, 10-12: mid blocks 0-2, 13-21: output blocks 0-8, 22: out` です。
222+
218223
`--network_args` で以下の引数を指定してください。
219224

220225
- `block_dims` : 各ブロックのdim (rank)を指定します。`"block_dims=2,2,2,2,4,4,4,4,6,6,6,6,8,6,6,6,6,4,4,4,4,2,2,2,2"` のように25個の数値を指定します。

docs/train_network_README-zh.md

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -101,6 +101,8 @@ LoRA的模型将会被保存在通过`--output_dir`选项指定的文件夹中
101101
* 当在Text Encoder相关的LoRA模块中使用与常规学习率(由`--learning_rate`选项指定)不同的学习率时,应指定此选项。可能最好将Text Encoder的学习率稍微降低(例如5e-5)。
102102
* `--network_args`
103103
* 可以指定多个参数。将在下面详细说明。
104+
* `--alpha_mask`
105+
* 使用图像的 Alpha 值作为遮罩。这在学习透明图像时使用。[PR #1223](https://github.com/kohya-ss/sd-scripts/pull/1223)
104106

105107
当未指定`--network_train_unet_only``--network_train_text_encoder_only`时(默认情况),将启用Text Encoder和U-Net的两个LoRA模块。
106108

fine_tune.py

Lines changed: 15 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -310,7 +310,11 @@ def fn_recursive_set_mem_eff(module: torch.nn.Module):
310310
init_kwargs["wandb"] = {"name": args.wandb_run_name}
311311
if args.log_tracker_config is not None:
312312
init_kwargs = toml.load(args.log_tracker_config)
313-
accelerator.init_trackers("finetuning" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
313+
accelerator.init_trackers(
314+
"finetuning" if args.log_tracker_name is None else args.log_tracker_name,
315+
config=train_util.get_sanitized_config_or_none(args),
316+
init_kwargs=init_kwargs,
317+
)
314318

315319
# For --sample_at_first
316320
train_util.sample_images(accelerator, args, 0, global_step, accelerator.device, vae, tokenizer, text_encoder, unet)
@@ -354,7 +358,9 @@ def fn_recursive_set_mem_eff(module: torch.nn.Module):
354358

355359
# Sample noise, sample a random timestep for each image, and add noise to the latents,
356360
# with noise offset and/or multires noise if specified
357-
noise, noisy_latents, timesteps, huber_c = train_util.get_noise_noisy_latents_and_timesteps(args, noise_scheduler, latents)
361+
noise, noisy_latents, timesteps, huber_c = train_util.get_noise_noisy_latents_and_timesteps(
362+
args, noise_scheduler, latents
363+
)
358364

359365
# Predict the noise residual
360366
with accelerator.autocast():
@@ -368,7 +374,9 @@ def fn_recursive_set_mem_eff(module: torch.nn.Module):
368374

369375
if args.min_snr_gamma or args.scale_v_pred_loss_like_noise_pred or args.debiased_estimation_loss:
370376
# do not mean over batch dimension for snr weight or scale v-pred loss
371-
loss = train_util.conditional_loss(noise_pred.float(), target.float(), reduction="none", loss_type=args.loss_type, huber_c=huber_c)
377+
loss = train_util.conditional_loss(
378+
noise_pred.float(), target.float(), reduction="none", loss_type=args.loss_type, huber_c=huber_c
379+
)
372380
loss = loss.mean([1, 2, 3])
373381

374382
if args.min_snr_gamma:
@@ -380,7 +388,9 @@ def fn_recursive_set_mem_eff(module: torch.nn.Module):
380388

381389
loss = loss.mean() # mean over batch dimension
382390
else:
383-
loss = train_util.conditional_loss(noise_pred.float(), target.float(), reduction="mean", loss_type=args.loss_type, huber_c=huber_c)
391+
loss = train_util.conditional_loss(
392+
noise_pred.float(), target.float(), reduction="mean", loss_type=args.loss_type, huber_c=huber_c
393+
)
384394

385395
accelerator.backward(loss)
386396
if accelerator.sync_gradients and args.max_grad_norm != 0.0:
@@ -471,7 +481,7 @@ def fn_recursive_set_mem_eff(module: torch.nn.Module):
471481

472482
accelerator.end_training()
473483

474-
if is_main_process and (args.save_state or args.save_state_on_train_end):
484+
if is_main_process and (args.save_state or args.save_state_on_train_end):
475485
train_util.save_state_on_train_end(args, accelerator)
476486

477487
del accelerator # この後メモリを使うのでこれは消す

finetune/prepare_buckets_latents.py

Lines changed: 27 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -11,15 +11,18 @@
1111

1212
import torch
1313
from library.device_utils import init_ipex, get_preferred_device
14+
1415
init_ipex()
1516

1617
from torchvision import transforms
1718

1819
import library.model_util as model_util
1920
import library.train_util as train_util
2021
from library.utils import setup_logging
22+
2123
setup_logging()
2224
import logging
25+
2326
logger = logging.getLogger(__name__)
2427

2528
DEVICE = get_preferred_device()
@@ -89,7 +92,9 @@ def main(args):
8992

9093
# bucketのサイズを計算する
9194
max_reso = tuple([int(t) for t in args.max_resolution.split(",")])
92-
assert len(max_reso) == 2, f"illegal resolution (not 'width,height') / 画像サイズに誤りがあります。'幅,高さ'で指定してください: {args.max_resolution}"
95+
assert (
96+
len(max_reso) == 2
97+
), f"illegal resolution (not 'width,height') / 画像サイズに誤りがあります。'幅,高さ'で指定してください: {args.max_resolution}"
9398

9499
bucket_manager = train_util.BucketManager(
95100
args.bucket_no_upscale, max_reso, args.min_bucket_reso, args.max_bucket_reso, args.bucket_reso_steps
@@ -107,7 +112,7 @@ def main(args):
107112
def process_batch(is_last):
108113
for bucket in bucket_manager.buckets:
109114
if (is_last and len(bucket) > 0) or len(bucket) >= args.batch_size:
110-
train_util.cache_batch_latents(vae, True, bucket, args.flip_aug, False)
115+
train_util.cache_batch_latents(vae, True, bucket, args.flip_aug, args.alpha_mask, False)
111116
bucket.clear()
112117

113118
# 読み込みの高速化のためにDataLoaderを使うオプション
@@ -208,7 +213,9 @@ def setup_parser() -> argparse.ArgumentParser:
208213
parser.add_argument("in_json", type=str, help="metadata file to input / 読み込むメタデータファイル")
209214
parser.add_argument("out_json", type=str, help="metadata file to output / メタデータファイル書き出し先")
210215
parser.add_argument("model_name_or_path", type=str, help="model name or path to encode latents / latentを取得するためのモデル")
211-
parser.add_argument("--v2", action="store_true", help="not used (for backward compatibility) / 使用されません(互換性のため残してあります)")
216+
parser.add_argument(
217+
"--v2", action="store_true", help="not used (for backward compatibility) / 使用されません(互換性のため残してあります)"
218+
)
212219
parser.add_argument("--batch_size", type=int, default=1, help="batch size in inference / 推論時のバッチサイズ")
213220
parser.add_argument(
214221
"--max_data_loader_n_workers",
@@ -231,18 +238,32 @@ def setup_parser() -> argparse.ArgumentParser:
231238
help="steps of resolution for buckets, divisible by 8 is recommended / bucketの解像度の単位、8で割り切れる値を推奨します",
232239
)
233240
parser.add_argument(
234-
"--bucket_no_upscale", action="store_true", help="make bucket for each image without upscaling / 画像を拡大せずbucketを作成します"
241+
"--bucket_no_upscale",
242+
action="store_true",
243+
help="make bucket for each image without upscaling / 画像を拡大せずbucketを作成します",
235244
)
236245
parser.add_argument(
237-
"--mixed_precision", type=str, default="no", choices=["no", "fp16", "bf16"], help="use mixed precision / 混合精度を使う場合、その精度"
246+
"--mixed_precision",
247+
type=str,
248+
default="no",
249+
choices=["no", "fp16", "bf16"],
250+
help="use mixed precision / 混合精度を使う場合、その精度",
238251
)
239252
parser.add_argument(
240253
"--full_path",
241254
action="store_true",
242255
help="use full path as image-key in metadata (supports multiple directories) / メタデータで画像キーをフルパスにする(複数の学習画像ディレクトリに対応)",
243256
)
244257
parser.add_argument(
245-
"--flip_aug", action="store_true", help="flip augmentation, save latents for flipped images / 左右反転した画像もlatentを取得、保存する"
258+
"--flip_aug",
259+
action="store_true",
260+
help="flip augmentation, save latents for flipped images / 左右反転した画像もlatentを取得、保存する",
261+
)
262+
parser.add_argument(
263+
"--alpha_mask",
264+
type=str,
265+
default="",
266+
help="save alpha mask for images for loss calculation / 損失計算用に画像のアルファマスクを保存する",
246267
)
247268
parser.add_argument(
248269
"--skip_existing",

finetune/tag_images_by_wd14_tagger.py

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -112,7 +112,6 @@ def main(args):
112112

113113
# モデルを読み込む
114114
if args.onnx:
115-
import torch
116115
import onnx
117116
import onnxruntime as ort
118117

0 commit comments

Comments
 (0)