PixArt-Sigma-1024px_512px-animetune

<ul><li>Pruned Model fp16 (1.15 GB): Inference checkpoint.</li><li>Training Data (2.69 GB): Fine-tuning diffusers model+OneTrainer config data.+10epoch_Inference checkpoint.</li><li>Config (12.48 KB): ComfyUI workflow. Due to upload requirements, it has been changed to .yml. Please change the extension to .json when using.</li><li>Pruned Model fp32 (2.29 GB):Full model. Unless you have a special reason, there is no need to use it as it is heavy.There is no difference in quality.</li><li>Pruned Model bf16 (1.61 GB):Experimental merge model v04. I have updated to my v0.2 model, and it feels better than before. I also changed the way I merge, so please be aware that the real style is strong. this model is very experimental, so it is a secret between you and me who found it... The extension has been changed to .bin for upload purposes. Please change it to .safetensors for use.I feel like the merge model creates a calmer atmosphere.It is fine to upload images generated with this merged model!However, if you merge this experimental merge model with other models, please do not upload the resulting merged model.This can lead to the dangerous practice of model inbreeding! Let's enjoy it within the scope of personal use.This request is only for the experimental merge model.It is okay to use my anime fine-tuning model as merge material to create merged models. You can upload them as well!</li></ul>■1024px_v0.1 has been updated.Tag training has become comparable to the 512px model.Anatomy has improved compared to before. There are still many flaws, though...However, I recently realized that instead of worrying about and fixing flaws, I find it more enjoyable to come up with prompts that create images with a good atmosphere. It's enjoyable to provide ChatGPT with a theme and have it generate natural language prompts!Generating with natural language is fun, but the accuracy might have decreased due to the progress in tag training. Therefore, it might be interesting to merge it with the previous model or other models fine-tuned with natural language.I am also sharing the training settings and test ComfyUI workflow.uses the "Uncond-Zero-for-ComfyUI" extension node, please install it if you encounter any errors.Please use the latest version as much as possible since some nodes used do not exist in the older versions of ComfyUI.this is not the definitive solution, so please try various workflows!For those who don’t want to waste time, I think it’s a good idea to practice with the 512px model first to see which tags work before using the 1024px model.This model has been trained for 12 epochs, but I also like the 10-epoch model because it is stable, so I am sharing it as well. The tag training is worse, but it might be better for natural language generation.Merging might also be interesting.An example of an interesting merging experiment: Perform a differential merge of my 512px model with a 512px base model to extract only the fine-tuning elements.Then, adding about "0.1-0.25" of the extracted elements to a 1024px model .There are more failures, but it's fun because it emphasizes the style.

pixartSigma1024px_1024pxV01.yml

pixartSigma1024px_1024pxV01.safetensors

pixartSigma1024px_1024pxV01_trainingData.zip

pixartSigma1024px_1024pxV01.bin

1024px_v0.1

1024px_v04

512px_0.7

512px_0.6

512px_v0.5

512px_v0.4

512px_v0.3

512px_v0.2

512px_v0.1

1024px_v0.31

1024px_v0.2

1024px_v0.0

512px_v0.0

Files with SHA256: c3a75be475236ed969ab63cf46f5666637a74bcad31a690634d0add1aad4bcfb