Hugging Face's logo

KBlueLeaf
/
DanTagGen-delta-rev2

Text Generation
Transformers
Safetensors
GGUF
English
llama
Not-For-All-Audiences
art
text-generation-inference
Inference Endpoints
Model card Files Community
1
DanTagGen-delta-rev2 / README.md
KBlueLeaf's picture
KBlueLeaf
Update README.md
9f559bc verified
preview code
|
raw
history blame contribute delete
3.16 kB
metadata
license: cc-by-sa-4.0
datasets:
  - KBlueLeaf/danbooru2023-sqlite
language:
  - en
library_name: transformers
pipeline_tag: text-generation
tags:
  - not-for-all-audiences
  - art
widget:
  - text: |-
      quality: masterpiece
      rating: safe
      artist: <|empty|>
      characters: <|empty|>
      copyrights: <|empty|>
      aspect ratio: 1.0
      target: <|short|>
      general: 1girl, solo, dragon girl, dragon horns, dragon tail<|input_end|>

DanTagGen - delta (rev2)

DanTagGen(Danbooru Tag Generator) is inspired from p1atdev's dart project. But with different arch, dataset, format and different training strategy.

Difference between versions

  • alpha: pretrain on 2M dataset, smaller batch size. Limited ability
  • beta: pretrain on 5.3M dataset, larger batch size. More stable, better ability with only a few information provided.
  • delta: pretrain on 7.2M dataset, larger batch size. Slightly underfit but better diversity. quality tag introduced.
    • rev2: resumed from delta, same dataset, 2 more epoch.

Model arch

This version of DTG is trained from scratch with 400M param LLaMA arch.(In my personal preference I will call it NanoLLaMA) Since it is llama arch. Theoritically it should be able to be used in any LLaMA inference interface.

This repo also provided converted FP16 gguf model and quantized 8bit/6bit gguf models. Basically it is recommended to use llama.cpp or llama-cpp-python to run this model. Which will be very fast.

Format 

prompt = f"""
quality: {quality or '<|empty|>'}
rating: {rating or '<|empty|>'}
artist: {artist.strip() or '<|empty|>'}
characters: {characters.strip() or '<|empty|>'}
copyrights: {copyrights.strip() or '<|empty|>'}
aspect ratio: {f"{aspect_ratio:.1f}" or '<|empty|>'}
target: {'<|' + target + '|>' if target else '<|long|>'}
general: {", ".join(special_tags)}, {general.strip().strip(",")}<|input_end|>
"""

for example:

quality: masterpiece
rating: safe
artist: <|empty|>
characters: <|empty|>
copyrights: <|empty|>
aspect ratio: 1.0
target: <|short|>
general: 1girl, solo, dragon girl, dragon horns, dragon tail<|input_end|>

And you may get something like:

rating: safe
artist: <|empty|>
characters: <|empty|>
copyrights: <|empty|>
aspect ratio: 1.0
target: <|short|>
general: 1girl, solo, dragon girl, dragon horns, dragon tail<|input_end|>open mouth, red eyes, long hair, pointy ears, tail, black hair, chinese clothes, simple background, dragon, hair between eyes, horns, china dress, dress, looking at viewer, breasts

Dataset and Training

I use the trainer I implemented in HakuPhi to run the training. with Total 12epoch on 7.2M data. This model have roughly 10~15B token seen.

The dataset is exported by HakuBooru with my danbooru sqlite database. Use the percentile of fav_count on each rating to filter the data. (2M = top 25%, 5.3M = top 75%)

Utilities