app.py

import torch
import gradio as gr
from optimum.onnxruntime import ORTModelForCausalLM
from transformers import AutoTokenizer

# https://huggingface.co/collections/p1atdev/dart-v2-danbooru-tags-transformer-v2-66291115701b6fe773399b0a
model_id = "p1atdev/dart-v2-sft"
model = ORTModelForCausalLM.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)
tokenizer_with_prefix_space = AutoTokenizer.from_pretrained(model_id, add_prefix_space=True)


# https://huggingface.co/docs/transformers/v4.44.2/en/internal/generation_utils#transformers.NoBadWordsLogitsProcessor
def get_tokens_as_list(word_list):
    "Converts a sequence of words into a list of tokens"
    tokens_list = []
    for word in word_list:
        tokenized_word = tokenizer_with_prefix_space([word], add_special_tokens=False).input_ids[0]
        tokens_list.append(tokenized_word)
    return tokens_list


def generate_tags(general_tags: str):
    # https://huggingface.co/p1atdev/dart-v2-sft#prompt-format
    general_tags = ",".join(tag.strip() for tag in general_tags.split(",") if tag)
    prompt = (
        "<|bos|>"
        # "<copyright></copyright>"
        # "<character></character>"
        "<|rating:general|><|aspect_ratio:tall|><|length:long|>"
        f"<general>{general_tags}<|identity:none|><|input_end|>"
    )

    inputs = tokenizer(prompt, return_tensors="pt").input_ids
    # bad_words_ids = get_tokens_as_list(word_list=[""])

    with torch.no_grad():
        outputs = model.generate(
            inputs,
            do_sample=True,
            temperature=1.0,
            top_p=1.0,
            top_k=100,
            max_new_tokens=128,
            num_beams=1,
            # bad_words_ids=bad_words_ids,
        )

    return ", ".join(
        [tag for tag in tokenizer.batch_decode(outputs[0], skip_special_tokens=True) if tag.strip() != ""]
    )


demo = gr.Interface(
    fn=generate_tags,
    inputs=gr.TextArea("1girl, black hair", lines=4),
    outputs=gr.Textbox(show_copy_button=True),
    clear_btn=None,
    analytics_enabled=False,
)

demo.launch()