accelerate launch ^
  --mixed_precision bf16 ^
  --num_cpu_threads_per_process 1 ^
  sd-scripts/flux_train_network.py ^
  --pretrained_model_name_or_path "E:\Pinokio\api\fluxgym.git\models\unet\bdsqlsz\flux1-dev2pro-single\flux1-dev2pro.safetensors" ^
  --clip_l "E:\Pinokio\api\fluxgym.git\models\clip\clip_l.safetensors" ^
  --t5xxl "E:\Pinokio\api\fluxgym.git\models\clip\t5xxl_fp16.safetensors" ^
  --ae "E:\Pinokio\api\fluxgym.git\models\vae\ae.sft" ^
  --cache_latents_to_disk ^
  --save_model_as safetensors ^
  --sdpa --persistent_data_loader_workers ^
  --max_data_loader_n_workers 2 ^
  --seed 42 ^
  --gradient_checkpointing ^
  --mixed_precision bf16 ^
  --save_precision bf16 ^
  --network_module networks.lora_flux ^
  --network_dim 4 ^
  --optimizer_type adamw8bit ^--sample_prompts="E:\Pinokio\api\fluxgym.git\outputs\davelandau-dev2pro-lora\sample_prompts.txt" --sample_every_n_steps="680" ^
  --learning_rate 8e-4 ^
  --cache_text_encoder_outputs ^
  --cache_text_encoder_outputs_to_disk ^
  --fp8_base ^
  --highvram ^
  --max_train_epochs 16 ^
  --save_every_n_epochs 4 ^
  --dataset_config "E:\Pinokio\api\fluxgym.git\outputs\davelandau-dev2pro-lora\dataset.toml" ^
  --output_dir "E:\Pinokio\api\fluxgym.git\outputs\davelandau-dev2pro-lora" ^
  --output_name davelandau-dev2pro-lora ^
  --timestep_sampling shift ^
  --discrete_flow_shift 3.1582 ^
  --model_prediction_type raw ^
  --guidance_scale 1 ^
  --loss_type l2 ^