File size: 2,127 Bytes
3dae8e7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
---
license: other
license_name: fair-ai-public-license-1.0-sd
license_link: https://freedevproject.org/faipl-1.0-sd/
datasets:
- pls2000/aiart_channel_nai3_geachu
base_model:
- OnomaAIResearch/Illustrious-xl-early-release-v0
tags:
- lora
---
# Lora Training (`arcain_2411.safetensors`)
Lora trained on Illustrious-xl v0.1, but this lora can applied with other ILXL-based models such as NoobAI-XL.
- Tool: kohya-ss/sd-scripts
- GPUs: 4x RTX3060
- Dataset: pls2000/aiart_channel_nai3_geachu + additional data until 24/11/14 - blue archive data
- Time taken: 50.5 hours (walltime)
#### lora_arcain.sh
```
NCCL_P2P_DISABLE=1 NCCL_IB_DISABLE=1 accelerate launch --num_cpu_threads_per_process 4 sdxl_train_network.py \
--network_train_unet_only \
--network_module="networks.lora" --network_dim 128 --network_alpha 128 \
--pretrained_model_name_or_path="/ai/data/sd/models/Stable-diffusion/SDXL/Illustrious-XL-v0.1.safetensors" \
--dataset_config="arcain.lora.toml" \
--output_dir="results/lora" --output_name="arcain-`date +%y%m`" \
--save_model_as="safetensors" \
--train_batch_size 2 --gradient_accumulation_steps 64 \
--learning_rate=1e-5 --optimizer_type="Lion8bit" \
--lr_scheduler="constant_with_warmup" --lr_warmup_steps 100 --optimizer_args "weight_decay=0.01" "betas=0.9,0.95" --min_snr_gamma 5 \
--sdpa \
--no_half_vae \
--cache_latents --cache_latents_to_disk \
--gradient_checkpointing \
--full_bf16 --mixed_precision="bf16" --save_precision="fp16" \
--ddp_timeout=10000000 \
--max_train_epochs 8 --save_every_n_epochs 1 \
--log_with wandb --log_tracker_name kohya-ss --wandb_run_name "arcain_`date +%y%m%d-%H%M`" --logging_dir wandb \
```
#### arcain.lora.toml
```
[general]
shuffle_caption = true
caption_tag_dropout_rate = 0.2
keep_tokens_separator = "|||"
caption_extension = ".txt"
[[datasets]]
enable_bucket = true
min_bucket_reso = 512
max_bucket_reso = 4096
resolution = 1024
[[datasets.subsets]]
image_dir = "/mnt/wd8tb/train/to_train"
num_repeats = 1
``` |