Version Arknights update:Arknights 版本使用50K张明日方舟图片训练。因为发现对画风造成了有趣影响，于是做成了大模型。基于原版pony训练。在A100 40G 上训练了 50小时。The Arknights version was trained using 50,000 images from Arknights.Due to the interesting effects observed on the art style, it was developed into a large model.Based on the original Pony model.Training took 50 hours on an A100 40G.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/3acb644a-78ff-4d97-8775-a4a28ceaf226/width=525/3acb644a-78ff-4d97-8775-a4a28ceaf226.jpeg" />Prompt<pre><code>score_9,score_8_up,score_7_up</code></pre>Negative prompt<pre><code>logo,score_4,score_5,score_6,lowres,(bad),text,error,fewer,extra,missing,worst quality,jpeg artifacts,low quality,watermark,unfinished,displeasing,oldest,early,chromatic aberration,signature,extra digits,artistic error,username,scan,[abstract],</code></pre>--Version Alpha update:-Further fine-tuning the furry style in the pony base model to a Japanese anime illustration style.-The current version of the model can be directly used for generating images.The base model remains chenkin_20w.safetensors trained using a 200,000 training dataset.Several manually selected high-quality illustrations are used for targeted training (provided by <a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">Euge</a>).Each illustration is of exceptional quality, labeled as "amazing."The trained model is merged with chenkin_20w.safetensors with a weight of 0.7.Therefore, the current quality indicators are:<pre><code>amazing,best，hight,score_9，</code></pre>or simply:<pre><code>amazing,best，hight,</code></pre>Alpha版本更新：-pony底模中的furry风格进一步微调为日系动漫插画风格。-现在版本的模型，可以直接用于出图。底模依然为使用20w训练集训练的chenkin_20w.safetensors对若干张人工挑选的超高质量插画进行针对训练。（由<a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">尤吉</a>提供）每一张都是千中挑一的水准，打上了`amazing` 的质量标签。训练后的模型，以0.7的权重与chenkin_20w.safetensors 合并。因此，现在的质量提示词是：<pre><code>amazing,best，hight,score_9，</code></pre>或者，仅仅使用：<pre><code>amazing,best，hight,</code></pre><hr />特别鸣谢：算力赞助 / GPU Sponsor: <a target="_blank" rel="ugc" href="https://neta.art/">Neta</a>特别鸣谢：算力赞助 / GPU Sponsor: <a target="_blank" rel="ugc" href="http://nieta.art">nieta.art</a><hr />This model may not be suitable for generating images and it is not recommended for beginners to download.Intended as a reference (or lesson) for colleagues who are training on Pony Diffusion V6 XL.Based on Pony Diffusion V6 XL, trained on 200k anime images, using an A40 48G for over 7 days.The model was trained to generate images in the style of Japanese anime, but did not achieve the expected results.1. chenkin_pony.safetensorsPretrained on 13k random anime images.2. chenkin_20w.safetensors (Current model)Trained on top of chenkin_pony.safetensors.Used 190k selected images from yande with no quality labels (provided by Miss Erity).Used 10k manually selected high-quality illustrations with the "hight" quality label (provided by <a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">Euge</a>).Using 100 manually selected ultra-high-quality illustrations with the "amazing" quality label (provided by <a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">Euge</a>).<pre><code>amazing,hight,score_9 </code></pre>Trained using the bmaltais/kohya_ss (<a target="_blank" rel="ugc" href="http://github.com">github.com</a>) project.This project consumed a significant amount of GPU computing power, thanks to <a target="_blank" rel="ugc" href="https://neta.art/">Neta</a> for providing the computing resources.Special thanks to GPU Sponsor: <a target="_blank" rel="ugc" href="https://neta.art/">Neta</a>.这个模型可能不适合用来生成图片，同时不建议初学者下载。旨在为更多在Pony Diffusion V6 XL上训练的同仁提供参考（或者说教训）基于Pony Diffusion V6 XL 训练，使用200k的动漫图片， 在 A40 48G 训练了超过7天。该模型是为了更好生成日本动漫风格的图片而训练，但是没有取得预期。1.chenkin_pony.safetensors使用了13k随机动漫图片进行预训练。2.chenkin_20w.safetensors （当前模型）在chenkin_pony.safetensors的基础上进行训练使用了190k张来自yande的精选图片，无质量标。（由二小姐提供）使用10k张人工挑选的高质量插画，打上了`hight` 的质量标签。（由<a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">尤吉</a>提供）使用100张人工挑选的超高质量插画，打上了`amazing` 的质量标签。（由<a target="_blank" rel="ugc" href="https://civarchive.com/user/Euge_">尤吉</a>提供）<pre><code>amazing,hight,score_9 </code></pre>使用<a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">bmaltais/kohya_ss (</a><a target="_blank" rel="ugc" href="http://github.com">github.com</a><a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">)</a> 项目进行训练。此项目耗费了大量的GPU的算力，感谢 <a target="_blank" rel="ugc" href="https://neta.art/">Neta </a>愿意为我提供算力。特别鸣谢：算力赞助 / GPU Sponsor: <a target="_blank" rel="ugc" href="https://neta.art/">Neta</a>训练参数如下（train set）：<pre><code>[sdxl_arguments]
cache_text_encoder_outputs = false
no_half_vae = false
min_timestep = 0
max_timestep = 1000

[model_arguments]
pretrained_model_name_or_path = "/root/autodl-tmp/stable-diffusion-webui/models/Stable-diffusion/chenkin_pony/chenkin_pony.safetensors"

[dataset_arguments]
shuffle_caption = true
debug_dataset = false
train_data_dir = "/root/autodl-tmp/20w"
dataset_repeats = 1
keep_tokens_separator = "|||"
resolution = "1024, 1024"
caption_dropout_rate = 0
caption_tag_dropout_rate = 0
caption_dropout_every_n_epochs = 0
token_warmup_min = 1
token_warmup_step = 0
enable_bucket = true
min_bucket_reso=640 
max_bucket_reso=2048
bucket_reso_steps=64


[training_arguments]
output_dir = "/root/autodl-tmp/stable-diffusion-webui/models/Stable-diffusion/chenkin_20w"
output_name = "chenkin_20w"
save_precision = "fp16"
train_batch_size=6 
vae_batch_size=4 
max_train_epochs=1 
save_every_n_steps = 2000
max_token_length = 225
mem_eff_attn = false
xformers = true
sdpa = false

max_data_loader_n_workers = 8
persistent_data_loader_workers = true
gradient_checkpointing = true
gradient_accumulation_steps = 1
mixed_precision = "fp16"

[sample_prompt_arguments]
sample_every_n_steps = 200
sample_sampler = "euler_a"
sample_prompts="/root/example.txt"

[saving_arguments]
save_model_as = "safetensors"

[optimizer_arguments]
optimizer_type = "AdaFactor"
learning_rate = 7.5e-7
train_text_encoder = false
learning_rate_te1 = 0
learning_rate_te2 = 0
optimizer_args = [ "scale_parameter=False", "relative_step=False", "warmup_init=False",]
lr_scheduler = "constant_with_warmup"
lr_warmup_steps = 100
max_grad_norm = 0</code></pre>

Files with SHA256: eaf641a05f97d0a9414b0fb320219435d68e3712cb56d0a5ff6e213da706690b