Closed DuroCuri closed 6 months ago
I open block lr, even though I open the last time I start training scripts, it still can not work. I can use these setting before.
15:42:19-451970 INFO accelerate launch --num_cpu_threads_per_process=4 "./train_network.py" --network_args "down_lr_weight=1,1,1.2,1.2,1,0.1,0.1,1,1,0.75,0.8,0.8" "mid_lr_weight=0.7" "up_lr_weight=1,1,1,1,1,1,1.2,1.2,1,1,1,0.7" --bucket_no_upscale --bucket_reso_steps=64 --cache_latents --caption_extension=".txt" --enable_bucket --min_bucket_reso=256 --max_bucket_reso=2048 --keep_tokens="1" --learning_rate="9.4e-05" --logging_dir="C:\Users\Administrator\Desktop\train\child\lele02\log" --lr_scheduler="cosine" --lr_scheduler_num_cycles="20" --lr_scheduler_power="0.3" --lr_warmup_steps="238" --max_data_loader_n_workers="0" --max_grad_norm="1" --resolution="1024" --max_train_steps="4760" --min_snr_gamma=5 --mixed_precision="bf16" --network_alpha="40" --network_args "preset=full" "conv_dim=16" "conv_alpha=10" "rank_dropout=0" "module_dropout=0" "factor=-1" "use_cp=False" "use_scalar=False" "decompose_both=False" "rank_dropout_scale=False" "algo=lokr" "train_norm=False" --network_dim=64 --network_module=lycoris.kohya --multires_noise_iterations="8" --multires_noise_discount="0.2" --optimizer_args "weight_decay=0.05" --optimizer_type="AdamW" --output_dir="C:/Users/Administrator/Desktop/train/child/lele02/output" --output_name="lele02" --pretrained_model_name_or_path="E:/Novelai/stable-diffusion-webui/models/Stable-diffusion/real istic/pureBeautyV1_pureBeautyV1.safetensors" --reg_data_dir="C:/Users/Administrator/Desktop/train/child/lele02/reg" --save_every_n_epochs="2" --save_model_as=safetensors --save_precision="bf16" --seed="1344" --train_batch_size="2" --train_data_dir="C:/Users/Administrator/Desktop/train/child/lele02/train" --v_pred_like_loss="0.0863" --xformers A matching Triton is not available, some optimizations will not be enabled. Error caught was: No module named 'triton' 2024-03-06 15:42:34 INFO prepare tokenizer train_util.py:3959 2024-03-06 15:42:35 INFO Using DreamBooth method. train_network.py:173 INFO prepare images. train_util.py:1469 INFO found directory train_util.py:1432 C:\Users\Administrator\Desktop\train\child\lele02\train\10_xiaotiancai girl contains 19 image files INFO found directory train_util.py:1432 C:\Users\Administrator\Desktop\train\child\lele02\train\6_xiaotiancai girl contains 8 image files INFO found directory train_util.py:1432 C:\Users\Administrator\Desktop\train\child\lele02\reg\1_girl contains 14 image files INFO 238 train images with repeating. train_util.py:1508 INFO 14 reg images. train_util.py:1511 INFO [Dataset 0] config_util.py:544 batch_size: 2 resolution: (1024, 1024) enable_bucket: True network_multiplier: 1.0 min_bucket_reso: 256 max_bucket_reso: 2048 bucket_reso_steps: 64 bucket_no_upscale: True
[Subset 0 of Dataset 0]
image_dir:
"C:\Users\Administrator\Desktop\train\child\lele02\train\10_xiaotiancai
girl"
image_count: 19
num_repeats: 10
shuffle_caption: False
keep_tokens: 1
keep_tokens_separator:
caption_dropout_rate: 0.0
caption_dropout_every_n_epoches: 0
caption_tag_dropout_rate: 0.0
caption_prefix: None
caption_suffix: None
color_aug: False
flip_aug: False
face_crop_aug_range: None
random_crop: False
token_warmup_min: 1,
token_warmup_step: 0,
is_reg: False
class_tokens: xiaotiancai girl
caption_extension: .txt
[Subset 1 of Dataset 0]
image_dir:
"C:\Users\Administrator\Desktop\train\child\lele02\train\6_xiaotiancai
girl"
image_count: 8
num_repeats: 6
shuffle_caption: False
keep_tokens: 1
keep_tokens_separator:
caption_dropout_rate: 0.0
caption_dropout_every_n_epoches: 0
caption_tag_dropout_rate: 0.0
caption_prefix: None
caption_suffix: None
color_aug: False
flip_aug: False
face_crop_aug_range: None
random_crop: False
token_warmup_min: 1,
token_warmup_step: 0,
is_reg: False
class_tokens: xiaotiancai girl
caption_extension: .txt
[Subset 2 of Dataset 0]
image_dir:
"C:\Users\Administrator\Desktop\train\child\lele02\reg\1_girl"
image_count: 14
num_repeats: 1
shuffle_caption: False
keep_tokens: 1
keep_tokens_separator:
caption_dropout_rate: 0.0
caption_dropout_every_n_epoches: 0
caption_tag_dropout_rate: 0.0
caption_prefix: None
caption_suffix: None
color_aug: False
flip_aug: False
face_crop_aug_range: None
random_crop: False
token_warmup_min: 1,
token_warmup_step: 0,
is_reg: True
class_tokens: girl
caption_extension: .txt
INFO [Dataset 0] config_util.py:550
INFO loading image sizes. train_util.py:794
100%|████████████████████████████████████████████████████████████████████████████████| 41/41 [00:00<00:00, 5108.17it/s]
INFO make buckets train_util.py:800
WARNING min_bucket_reso and max_bucket_reso are ignored if bucket_no_upscale is train_util.py:817
set, because bucket reso is defined by image size automatically /
bucket_no_upscaleが指定された場合は、bucketの解像度は画像サイズから自動計
算されるため、min_bucket_resoとmax_bucket_resoは無視されます
INFO number of images (including repeats) / train_util.py:846
各bucketの画像枚数(繰り返し回数を含む)
INFO bucket 0: resolution (512, 768), count: 170 train_util.py:851
INFO bucket 1: resolution (768, 768), count: 72 train_util.py:851
INFO bucket 2: resolution (768, 1152), count: 30 train_util.py:851
INFO bucket 3: resolution (832, 1152), count: 68 train_util.py:851
INFO bucket 4: resolution (1024, 1024), count: 36 train_util.py:851
INFO bucket 5: resolution (1152, 768), count: 100 train_util.py:851
INFO mean ar error (without repeats): 0.025701847271969253 train_util.py:856
INFO preparing accelerator train_network.py:226
accelerator device: cuda
INFO loading model for process 0/1 train_util.py:4111
INFO load StableDiffusion checkpoint: train_util.py:4066
E:/Novelai/stable-diffusion-webui/models/Stable-diffusion/realistic/pure
BeautyV1_pureBeautyV1.safetensors
INFO UNet2DConditionModel: 64, 8, 768, False, False original_unet.py:1387
2024-03-06 15:42:38 INFO loading u-net:
Can you share your .json config file? I will try to load and use it. I tried to reproduce the issue on my system using the latest code version and all work fine...
lvbu_test_20240306-132245.json This file, I reset my version, now it can runs.
OK, so it was a setting issue on your system. It happen.
My current version is commit 62fbae6b3eedffbefd5964dd5c60c492ae905fc7 (HEAD -> master, tag: v22.6.0) Merge: bfe8b06 36b666b Author: bmaltais bernard@ducourier.com Date: Sat Jan 27 14:03:03 2024 -0500
Merge pull request #1907 from bmaltais/dev
v22.6.0
Just git pull to newest version, this bug comes again. So it happens because my system setting? It comes out with AttributeError: 'LokrModule' object has no attribute 'lokr_w1'. Did you mean: 'lokr_w1_a'? I can not find anywhere to set up it.
This appear to be a bug with the LyCORIS code... something that would need to be investigated by the author... Not something that is configured as part of the GUI... I updated the requirements to use the latest version og LyCORIS... probably a bug introduced in it.
Thank you sincerely for your invaluable help and guidance!