<@205854764540362752> Hello Dr. Furkan Gözükara, i'm doing all with your video instruction for windows 11 https://www.youtube.com/watch?v=FvpWy1x5etM, and i don't understand why it doesn't work properly, i have 4090RTX and 64gb RAM and have very slow training, i tried for finetune configs and for training, ssame situation:
enable full bf16 training.
running training / 学習開始
num examples / サンプル数: 115
num batches per epoch / 1epochのバッチ数: 29
num epochs / epoch数: 40
batch size per device / バッチサイズ: 4
gradient accumulation steps / 勾配を合計するステップ数 = 1
total optimization steps / 学習ステップ数: 1150
steps: 0%| | 0/1150 [00:00<?, ?it/s]
epoch 1/40
2025-03-02 18:58:34 INFO epoch is incremented. current_epoch: 0, epoch: 1 train_util.py:764
C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\lib\site-packages\torch\autograd\graph.py:825: UserWarning: cuDNN SDPA backward got grad_output.strides() != output.strides(), attempting to materialize a grad_output with matching strides... (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\cudnn\MHA.cpp:676.)
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
steps: 0%| | 1/1150 [30:36<586:07:19, 1836.41s/it, avr_loss=0.451]

GPU is loading for 100% but only 35W from 500W
Almost 30 min to epoch, i uxecuted all scripts installed all like in tutorials. 

This logs when using for training finetune conf for 16GB:
Starting the GUI... this might take some time...
19:49:59-213886 WARNING  Skipping requirements verification.
19:49:59-215887 INFO     headless: False
19:49:59-216887 INFO     Using shell=True when running external commands...
* Running on local URL:  http://127.0.0.1:7860

To create a public link, set `share=True` in `launch()`.
19:50:25-273788 INFO     Loading config...
19:51:28-840855 ERROR    Instance prompt missing...
19:52:20-316493 INFO     Start training Dreambooth...
19:52:20-318492 INFO     Validating lr scheduler arguments...
19:52:20-319493 INFO     Validating optimizer arguments...
19:52:20-320493 INFO     Validating C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\model existence and
                         writability... SUCCESS
19:52:20-322495 INFO     Validating C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/flux1-dev.safetensors
                         existence... SUCCESS
19:52:20-322495 INFO     Validating C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\img existence...
                         SUCCESS
19:52:20-323493 INFO     Folder 1_ohwx man: 1 repeats found
19:52:20-326493 INFO     Folder 1_ohwx man: 115 images found
19:52:20-327494 INFO     Folder 1_ohwx man: 115 * 1 = 115 steps
19:52:20-328495 INFO     Regularization factor: 1
19:52:20-329494 INFO     Total steps: 115
19:52:20-330494 INFO     Train batch size: 1
19:52:20-330494 INFO     Gradient accumulation steps: 1
19:52:20-331494 INFO     Epoch: 20
19:52:20-332494 INFO     max_train_steps (115 / 1 / 1 * 20 * 1) = 2300
19:52:20-334494 INFO     lr_warmup_steps = 0
19:52:20-337493 INFO     Saving training config to
                         C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\model\Quality_1_20250302-195220
                         .json...
19:52:20-340493 INFO     Executing command:
                         C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\Scripts\accelerate.EXE
                         launch --dynamo_backend no --dynamo_mode default --gpu_ids 0 --mixed_precision bf16
                         --num_processes 1 --num_machines 1 --num_cpu_threads_per_process 2
                         C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/kohya_ss/sd-scripts/flux_train.py
                         --config_file
                         C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\model/config_dreambooth-2025030
                         2-195220.toml
C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\lib\site-packages\diffusers\utils\outputs.py:63: FutureWarning: `torch.utils._pytree._register_pytree_node` is deprecated. Please use `torch.utils._pytree.register_pytree_node` instead.
  torch.utils._pytree._register_pytree_node(
C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\lib\site-packages\diffusers\utils\outputs.py:63: FutureWarning: `torch.utils._pytree._register_pytree_node` is deprecated. Please use `torch.utils._pytree.register_pytree_node` instead.
  torch.utils._pytree._register_pytree_node(
2025-03-02 19:52:27 INFO     Loading settings from                                                    train_util.py:4625
                             C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\model/co
                             nfig_dreambooth-20250302-195220.toml...
2025-03-02 19:52:27 INFO     Using DreamBooth method.                                                  flux_train.py:115
                    INFO     prepare images.                                                          train_util.py:2053
                    INFO     get image size from name of cache files                                  train_util.py:1944
100%|████████████████████████████████████████████████████████████████████████████| 115/115 [00:00<00:00, 114980.92it/s]
                    INFO     set image size from cache files: 115/115                                 train_util.py:1974
                    INFO     found directory                                                          train_util.py:2000
                             C:\Users\PC\Desktop\AI\Projects\SwarmUI\Models\diffusion_models\img\1_oh
                             wx man contains 115 image files
read caption: 100%|███████████████████████████████████████████████████████████████| 115/115 [00:00<00:00, 19167.29it/s]
                    INFO     115 train images with repeats.                                           train_util.py:2096
2025-03-02 19:52:28 INFO     0 reg images with repeats.                                               train_util.py:2100
                    WARNING  no regularization images / 正則化画像が見つかりませんでした              train_util.py:2105
                    INFO     [Dataset 0]                                                              config_util.py:575
                               batch_size: 1
                               resolution: (1024, 1024)
                               enable_bucket: False

                               [Subset 0 of Dataset 0]
                                 image_dir:
                             "C:\Users\PC\Desktop\AI\Projects\SwarmUI\Models\diffusion_models\img\1_o
                             hwx man"
                                 image_count: 115
                                 num_repeats: 1
                                 shuffle_caption: False
                                 keep_tokens: 0
                                 caption_dropout_rate: 0
                                 caption_dropout_every_n_epochs: 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,
                                 alpha_mask: False
                                 custom_attributes: {}
                                 is_reg: False
                                 class_tokens: ohwx man
                                 caption_extension: .txt


                    INFO     [Prepare dataset 0]                                                      config_util.py:587
                    INFO     loading image sizes.                                                      train_util.py:972
100%|████████████████████████████████████████████████████████████████████████████| 115/115 [00:00<00:00, 115008.34it/s]
                    INFO     prepare dataset                                                           train_util.py:997
                    INFO     Checking the state dict: Diffusers or BFL, dev or schnell                  flux_utils.py:43
                    INFO     prepare accelerator                                                       flux_train.py:186
accelerator device: cuda
                    INFO     Building AutoEncoder                                                      flux_utils.py:144
                    INFO     Loading state dict from                                                   flux_utils.py:149
                             C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/ae.safetensors
                    INFO     Loaded AE: <All keys matched successfully>                                flux_utils.py:152
                    INFO     [Dataset 0]                                                              train_util.py:2589
                    INFO     caching latents with caching strategy.                                   train_util.py:1097
                    INFO     caching latents...                                                       train_util.py:1146
100%|█████████████████████████████████████████████████████████████████████████████| 115/115 [00:00<00:00, 10454.85it/s]
C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\lib\site-packages\transformers\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
  warnings.warn(
You are using the default legacy behaviour of the <class 'transformers.models.t5.tokenization_t5.T5Tokenizer'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thoroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565
2025-03-02 19:52:30 INFO     Building CLIP-L                                                           flux_utils.py:179
                    INFO     Loading state dict from                                                   flux_utils.py:275
                             C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/clip_l.safetens
                             ors
                    INFO     Loaded CLIP-L: <All keys matched successfully>                            flux_utils.py:278
                    INFO     Loading state dict from                                                   flux_utils.py:330
                             C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/t5xxl_fp16.safe
                             tensors
2025-03-02 19:52:53 INFO     Loaded T5xxl: <All keys matched successfully>                             flux_utils.py:333
2025-03-02 19:52:55 INFO     [Dataset 0]                                                              train_util.py:2611
                    INFO     caching Text Encoder outputs with caching strategy.                      train_util.py:1280
                    INFO     checking cache validity...                                               train_util.py:1291
100%|██████████████████████████████████████████████████████████████████████████████| 115/115 [00:00<00:00, 6051.40it/s]
                    INFO     no Text Encoder outputs to cache                                         train_util.py:1318
                    INFO     cache Text Encoder outputs for sample prompt:                             flux_train.py:249
                             C:/Users/PC/Desktop/AI/Projects/SwarmUI/Models/diffusion_models\model\sam
                             ple/prompt.txt
                    INFO     Checking the state dict: Diffusers or BFL, dev or schnell                  flux_utils.py:43
                    INFO     Building Flux model dev from BFL checkpoint                               flux_utils.py:101
                    INFO     Loading state dict from                                                   flux_utils.py:118
                             C:/Users/PC/Desktop/AI/Projects/Kohya_FLUX_DreamBooth_v17/flux1-dev.safet
                             ensors
                    INFO     Loaded Flux: <All keys matched successfully>                              flux_utils.py:137
FLUX: Gradient checkpointing enabled. CPU offload: False
                    INFO     enable block swap: blocks_to_swap=22                                      flux_train.py:304
FLUX: Block swap enabled. Swapping 22 blocks, double blocks: 11, single blocks: 22.
number of trainable parameters: 11901408320
prepare optimizer, data loader etc.
                    INFO     use Adafactor optimizer | {'scale_parameter': False, 'relative_step':    train_util.py:4937
                             False, 'warmup_init': False, 'weight_decay': 0.01}
                    WARNING  because max_grad_norm is set, clip_grad_norm is enabled. consider set to train_util.py:4965
                             0 /
                             max_grad_normが設定されているためclip_grad_normが有効になります。0に設定
                             して無効にしたほうがいいかもしれません
                    WARNING  constant_with_warmup will be good /                                      train_util.py:4969
                             スケジューラはconstant_with_warmupが良いかもしれません
enable full bf16 training.
running training / 学習開始
  num examples / サンプル数: 115
  num batches per epoch / 1epochのバッチ数: 115
  num epochs / epoch数: 20
  batch size per device / バッチサイズ: 1
  gradient accumulation steps / 勾配を合計するステップ数 = 1
  total optimization steps / 学習ステップ数: 2300
steps:   0%|                                                                                  | 0/2300 [00:00<?, ?it/s]
epoch 1/20
2025-03-02 19:53:25 INFO     epoch is incremented. current_epoch: 0, epoch: 1                          train_util.py:764
C:\Users\PC\Desktop\AI\Projects\Kohya_FLUX_DreamBooth_v17\kohya_ss\venv\lib\site-packages\torch\autograd\graph.py:825: UserWarning: cuDNN SDPA backward got grad_output.strides() != output.strides(), attempting to materialize a grad_output with matching strides... (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\cudnn\MHA.cpp:676.)
  return Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
steps:   4%|██▎                                                    | 97/2300 [11:16<4:16:08,  6.98s/it, avr_loss=0.489]

Used AI for this problem and don't understand how to win this situation ))) If you can help it will bery good, thank you