点滴 intravenous drip-XLPony

#点滴 v001

### 명령줄 설정

- accelerate.EXE launch --dynamo_backend no --dynamo_mode default --mixed_precision bf16 --num_processes 2 --num_machines 1 --num_cpu_threads_per_process 2

- sdxl_train_network.py --config_file config_lora-20240618-184456.toml

- --lr_scheduler_type CosineAnnealingLR

- --lr_scheduler_args T_max=1040 eta_min=1e-7

- --network_args loraplus_lr_ratio=16

- --alpha_mask

### toml 설정 파일

- bucket_no_upscale = true

- bucket_reso_steps = 64

- cache_latents = true

- cache_latents_to_disk = true

- caption_dropout_rate = 0.05

- caption_extension = ".txt"

- clip_skip = 1

- dynamo_backend = "no"

- enable_bucket = true

- epoch = 100

- gradient_accumulation_steps = 1

- gradient_checkpointing = true

- huber_c = 0.1

- huber_schedule = "snr"

- keep_tokens = 1

- learning_rate = 0.0001

- log_with = "wandb"

- logging_dir = "log"

- loss_type = "smooth_l1"

- lr_scheduler = "cosine"

- lr_scheduler_args = []

- lr_scheduler_num_cycles = 1

- lr_scheduler_power = 1

- max_bucket_reso = 2048

- max_data_loader_n_workers = 0

- max_grad_norm = 0.1

- max_timestep = 1000

- max_token_length = 225

- max_train_epochs = 100

- max_train_steps = 10400

- mem_eff_attn = true

- min_bucket_reso = 256

- min_snr_gamma = 5

- mixed_precision = "bf16"

- network_alpha = 32

- network_args = ["rank_dropout=0.2", "module_dropout=0.2"]

- network_dim = 32

- network_dropout = 0.2

- network_module = "networks.lora"

- no_half_vae = true

- noise_offset_type = "Original"

- optimizer_args = ["weight_decay=0.01"]

- optimizer_type = "AdamW8bit"

- output_dir = "output"

- output_name = "intravenousdrip-XLPony_001"

- pretrained_model_name_or_path = "ponyDiffusionV6XL_v6StartWithThisOne.safetensors"

- prior_loss_weight = 0.1

- resolution = "1024,1024"

- sample_prompts = "prompt.txt"

- sample_sampler = "euler_a"

- save_every_n_epochs = 10

- save_model_as = "safetensors"

- save_precision = "bf16"

- save_state = true

- seed = 1

- shuffle_caption = true

- text_encoder_lr = 1e-5

- train_batch_size = 1

- unet_lr = 0.0001

- xformers = true