NeoLoger
commented
1 year ago same exact issue, it started after the latest update I think.
Replace CrossAttention.forward to use xformers
caching latents.
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 498/498 [00:29<00:00, 17.12it/s]
import network module: networks.lora
create LoRA for Text Encoder: 72 modules.
create LoRA for U-Net: 192 modules.
enable LoRA for text encoder
enable LoRA for U-Net
prepare optimizer, data loader etc.
Traceback (most recent call last):
File "S:\kohya\kohya_ss\train_network.py", line 507, in <module>
train(args)
File "S:\kohya\kohya_ss\train_network.py", line 150, in train
optimizer_name, optimizer_args, optimizer = train_util.get_optimizer(args, trainable_params)
File "S:\kohya\kohya_ss\library\train_util.py", line 1536, in get_optimizer
assert optimizer_type is None or optimizer_type == "", "both option use_8bit_adam and optimizer_type are specified / use_8bit_adamとoptimizer_typeの両方のオプションが 指定されています"
AssertionError: both option use_8bit_adam and optimizer_type are specified / use_8bit_adamとoptimizer_typeの両方のオプションが指定されています
Traceback (most recent call last):
File "C:\Users\Yvggeniy\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\Yvggeniy\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 86, in _run_code
exec(code, run_globals)
File "S:\kohya\kohya_ss\venv\Scripts\accelerate.exe\__main__.py", line 7, in <module>
File "S:\kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 45, in main
args.func(args)
File "S:\kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 1104, in launch_command
simple_launcher(args)
File "S:\kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 567, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['S:\\kohya\\kohya_ss\\venv\\Scripts\\python.exe', 'train_network.py', '--pretrained_model_name_or_path=S:/stable-diffusion/stable-diffusion-webui/models/Stable-diffusion/My_mixes/My_anyhentai_abyss_reva1_grape32_pov.safetensors', '--train_data_dir=V:/ImagesForSDTrining/Lora training/all the way through/image', '--resolution=512,512', '--output_dir=V:/ImagesForSDTrining/Lora training/all the way through/model', '--logging_dir=V:/ImagesForSDTrining/Lora training/all the way through/log', '--network_alpha=128', '--training_comment=Trained on My_anyhentai_abyss_reva1_grape32_pov', '--save_model_as=safetensors', '--network_module=networks.lora', '--text_encoder_lr=5e-5', '--unet_lr=0.0001', '--network_dim=128', '--output_name=all the way through_v3', '--lr_scheduler_num_cycles=1', '--learning_rate=0.0001', '--lr_scheduler=constant', '--train_batch_size=2', '--max_train_steps=24900', '--save_every_n_epochs=1', '--mixed_precision=bf16', '--save_precision=bf16', '--seed=1234', '--caption_extension=.txt', '--cache_latents', '--optimizer_type=AdamW', '--max_data_loader_n_workers=1', '--clip_skip=2', '--bucket_reso_steps=64', '--xformers', '--use_8bit_adam', '--bucket_no_upscale']' returned non-zero exit status 1.
Throvtek
Rakshesha2024
bmaltais
jiuxiaojian
JasonTaverner
2blackbar
gekkomode
mxharms
ludwigjer
StudioCMC
mtnmecca
SwannSchilling
Ekaitza1985
mykeehu
udterry
Ahoo-Mun
noobbuddy
Havasiz
I started the latest version with the usual parameters, but now I got an error. Seems there is something wrong with the optimizer? I tried, with AdamW and AdamW8bit, without success.
If I turned off the "Use 8bit adam" option and selected AdamW, it started. So either this option should be disabled or, in the case of AdamW, ignore it when compiling the command.
What is the difference between AdamW and AdamW8bit? If I choose the former, will it cause a burn-in (as if I've overtrain it)?