app.py

import torch
torch.set_grad_enabled(False)
import gradio as gr
from datetime import datetime
from inference_tagger_standalone import *
# from huggingface_hub import hf_hub_download
# hf_hub_download(repo_id="lodestones/tagger-experiment", filename="tagger_proto.safetensors", local_dir=".")
import os
os.system("wget -nv https://huggingface.co/lodestones/tagger-experiment/resolve/main/tagger_proto.safetensors")

model = Tagger(checkpoint_path="./tagger_proto.safetensors", vocab_path="./tagger_vocab.json", max_size=1024)

def get_tags(image, threshold, top_k):
    current_datetime = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
    print(f"{current_datetime}: started.")
    results = model.predict(image, topk=top_k, threshold=threshold)
    temp = []
    return_dict = dict()
    for rank, (tag, score) in enumerate(results, 1):
        return_dict[tag] = score
        temp.append(tag.replace(" ", "_"))
    return_str = " ".join(temp)
    current_datetime = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
    print(f"{current_datetime}: finished.\n")
    return return_str, return_dict

demo = gr.Interface(
    get_tags,
    inputs=[
        gr.Image(label="Source", sources=['upload',], type='filepath'), 
        gr.Slider(minimum=0.00, maximum=1.00, step=0.01, value=0.30, label="Threshold"),
        gr.Slider(minimum=0, maximum=500, step=1, value=30, label="Top K")
    ],
    outputs=[
        gr.Textbox(label="Tag String"),
        gr.Label(label="Tag Predictions", num_top_classes=200),
    ],
)

demo.launch(ssr_mode=False)