途中で送ってしもうた
帰宅したやで
>>419,>>449ニキたちを参考に
venv\Scripts\activate　pip install upgrade transformers　pip install sentencepiece
学習実行に--split_modeを外し
accelerate launch --num_cpu_threads_per_process 1 flux_train_network.py --pretrained_model_name_or_path "C:\models\Stable-diffusion\flux1-dev.safetensors" --train_data_dir "F:\AI\loratest4noa\sozai" --output_dir "F:\AI\loratest4noa\output" --network_module "networks.lora_flux" --sdpa --gradient_checkpointing --persistent_data_loader_workers --cache_latents --cache_latents_to_disk --max_data_loader_n_workers 1 --enable_bucket --save_model_as "safetensors" --lr_scheduler_num_cycles 4 --mixed_precision "bf16" --learning_rate 0.0002 --resolution 512 --train_batch_size 1 --max_train_epochs 6 --network_dim 16 --network_alpha 8 --optimizer_type "Lion" --output_name "test" --ae
"H:\VAE\ae.safetensors" --save_precision "fp16" --lr_scheduler "cosine_with_restarts" --min_bucket_reso 320 --max_bucket_reso 1536 --caption_extension ".txt" --seed 42 --network_train_unet_only --cache_text_encoder_outputs --cache_text_encoder_outputs_to_disk --fp8_base --loss_type "l2" --clip_l "H:\CLIP\clip_l.safetensors" --t5xxl "H:\flux2\models\TE\t5xxl_fp16.safetensors" --network_args "train_blocks=single"

んでこれやな(変わらず)
https://files.catbox.moe/6yfliy.png

うーーーーーん