KBlueLeaf

/

DanTagGen-delta-rev2

like 26

This mode is heavily depends on special workflow and format
it is recommended to use my Library to inference on this model
https://github.com/KohakuBlueleaf/KGen

KGen support both transformers model or llama-cpp-python

This mode is heavily depends on special workflow and format
it is recommended to use my Library to inference on this model
https://github.com/KohakuBlueleaf/KGen

KGen support both transformers model or llama-cpp-python

@KBlueLeaf
In addition, I'm trying to import this into Ollama. I'll let you know how it goes!

Thanks!
Peter

I imported the f16 gguf model into ollama:

1. create Modelfile, with 1 line in it, declaring the local gguf model path
2. ollama create
3. ollama run

This is my test output:

ollama run danTagGen
>>> """
... 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|>
... """
general:p, blush stickers, looking at viewer, blue hair, tongue out, :q, purple eyes, tail, tongue, horns, short hair, chibi

>>> Send a message (/? for help)

The result is not quite ideal, maybe because that I didn't import the tokenizer. Would you mind look into that a bit?@KBlueLeaf Thanks!

Best,
Peter

No new line after input end

Also try to adjust the temperature