PixArt-Sigma-1024px_512px-animetune

<ul><li>Pruned Model fp16 (1.15 GB): Inference checkpoint.</li><li>Pruned Model bf16 (9.99 GB): Fine-tuning diffusers model+OneTrainer config data.</li><li>Training Data (68.78 KB): ComfyUI workflow. 12/22 update!I have added quality prompts. This should help maintain high quality even with fewer tags. The previous workflows are also included in the folder. 12/27 I have also added the TIPO workflow. It reduces the effort of crafting prompts and allows for easily generating high-quality images, so I highly recommend it.</li></ul>■ 10/27 v0.3 → v0.31 update<ul><li>Trained for 2 additional epochs=+25,000step. The model has further learned the concepts.</li><li>update Inference checkpoint&amp;Training Data.I have backed up the checkpoints from the past 30 epochs within the training data.</li><li>There are no major changes. Since generating sample images takes a lot of effort, I made this a minor update. I apologize for not announcing this to those who have already downloaded the previous version. I understand that this may come across as an insincere response, but I would appreciate your understanding.</li></ul>●We expanded the dataset to 400,000 images and trained for 8 epochs. (v0.3 is 6ep)I believe it has learned a wider range of concepts compared to before.While the diversity has increased, in some cases, aesthetic images were easier to generate with previous models, as they had more consistency in style.●The anatomy of hands and feet is still challenging. Instead of focusing too much on fixing it, I think it's better to enjoy the overall atmosphere.It might also be a good idea to use i2i with other models like SDXL or SD1.5 to correct anatomy and details.●It is recommended to use as specific tags or natural language prompts as possible with this model.Short prompts like "1girl,beach" may not produce very good images.●There seems to be a pattern in which tags or prompts improve quality.Tags that consistently have high-quality images on Danbooru might help increase image quality.●If you don’t want to waste time, it might be a good idea to use the 512px model first to practice which prompts are effective.●Also, the model has not been trained on quality tags or negative prompts.It has not been trained on images with potentially harmful effects, like sketche or monochrome images.However, all 400,000 images are of high quality, so there is a possibility that any tag could improve quality. The more tags, the better.●If you can't come up with a prompt, try using the prompt auto-generation below.<a target="_blank" rel="ugc" href="https://huggingface.co/spaces/KBlueLeaf/TIPO-DEMO">https://huggingface.co/spaces/KBlueLeaf/TIPO-DEMO</a>●There are likely tags that can be used as negative prompts.Tags related to styles, like the ones listed below, can be included in negative prompts to switch to a different style.1980s (style)-1990s (style),00s,10s,simple background,anime screencap, realistic,figure etc...●Natural language prompts can also be enjoyable.When I want to create beautiful images, I use natural language.When I want to use concepts that the base model doesn't know, I use tags.You can also combine the strengths of both by adding tags into natural language prompts when appropriate.●I’m using only general tags and character names. Tags like work titles, artist names, highres, official art, game CG, and other meta tags have been removed. Exceptionally, 'anime screenshot' has been individually trained. These tags were removed because they were commonly applied to many unrelated images and could overload the tokens, causing the model to learn unwanted concepts. If you have any specific requests, I can prepare a custom dataset for the desired concepts.■Breakdown of the dataset:Existing: 70,000Concept enhancement: 200,000Aesthetic enhancement: 130,000●Pruned Model bf16 (2.13 GB):Experimental merge model.this model is very experimental, so it is a secret between you and me found it...The extension has been changed to .bin for upload purposes. Please change it to .zip for use.Please check the specific notes included in the extracted zip folder.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!Also, this merged model is broken due to excessive additive merging. Let's enjoy it within the scope of personal use.

pixartSigma1024px_1024pxV031.zip

pixartSigma1024px_1024pxV03.safetensors

pixartSigma1024px_1024pxV031.safetensors

pixartSigma1024px_1024pxV031.bin

pixartSigma1024px_1024pxV031_trainingData.zip

pixartSigma1024px_1024pxV031.sft

1024px_v0.31

1024px_v04

512px_0.7

512px_0.6

512px_v0.5

512px_v0.4

512px_v0.3

512px_v0.2

512px_v0.1

1024px_v0.2

1024px_v0.1

1024px_v0.0

512px_v0.0

Files with SHA256: 73333ec4d02a958d0a42597df992c0eff99d8c17b39252b4e9eed85ea5e92326