DraconicDragon

import os
# 禁用matplotlib的GUI集成，避免VSCode调试器冲突
os.environ['MPLBACKEND'] = 'Agg'

import tkinter as tk
from tkinter import ttk, filedialog, messagebox
import cv2
from PIL import Image, ImageTk
import threading
from pathlib import Path

DraconicDragon

import argparse
import gzip
import logging
from pathlib import Path

import piexif
from PIL import Image

# ─── Configuration ─────────────────────────────────────────────────────────────
EXTRACT_STEALTH_METADATA = False  # Extract stealth metadata from embedded bits

DraconicDragon

# Before running the script, put it in the directory you want it to convert PNGs into WebPs
# Then run it, it will go through every image in the same directory the script is in (no single image mode sorry, too lazy)
# and save it in a subdirectory. It will skip any existing WebPs so if you want to replace the existing webps with new ones
# like with stealth enabled this time, if you had it disabled before, then youll have to delete the existing webps.

# There are some boolean values you can change as well as some other stuff in the "Configuration" part
# like a save as animated webp option which is useful for sharing the webp with a1111 style metadata in usercomment over discord
# because discord won't remove the metadata if it detects the webp as animated webp

# What this script doesn't do:

DraconicDragon

from transformers import CLIPTokenizer

# Load the tokenizer from the local directory
tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")

# Your input text
text = input("Input text:\n")

# Tokenize the text
tokens = tokenizer(text, return_tensors="pt")

DraconicDragon

# Define path
$downloadsPath = "$env:USERPROFILE\Downloads"
$user = "$env:USERNAME"
$identity = "$env:USERDOMAIN\$user"

# Get current ACL
$acl = Get-Acl $downloadsPath

# Remove all existing rules to start clean
$acl.Access | ForEach-Object { $acl.RemoveAccessRule($_) }

DraconicDragon

import requests
import torch
from PIL import Image
from transformers import AutoModelForCausalLM, AutoProcessor

device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

model_id = r"microsoft/Florence-2-base" # can also use path/to/huggingface/hub/florence2-repo-name/snapshots/hash or something like that
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch_dtype, trust_remote_code=True).to(device)

DraconicDragon

import sys
import re
class GreenOutput:
    # regex pattern to match URLs, IP and ports
    url_pattern = re.compile(
        r'((http|https|ftp|ftps):\/\/'  # protocol
        r'(\d{1,3}\.){3}\d{1,3}'        # ip
        r'(?::\d+)?'                    # port number if existing
        r'(?:[\/\w.,@?^=%&:/~+#-]*)?'   # rest of URL
        r'|'                            # OR

DraconicDragon

import os
import tkinter as tk
from tkinter import filedialog

import piexif
import piexif.helper
from PIL import Image


class PNGtoJPGConverterApp:

DraconicDragon

:root {
    --sidebar-w-collapsed: 36px; /*used for  collapsable sidebar*/
    --sidebar-h-extended: 240px;
    --memberlist-collapsed: 62px;
    --memberlist-expanded: 240px;
    --memberlist-transition: 0.35s ease-in-out;
}

/*vc text is correct with channel sidebar and main chat margin*/
.chat_f8f70f { margin-left: 0px !important; }

DraconicDragon

XAML Chessboard with Gridview, ItemsPanel, ItemsTemplate

XAML

<GridView
    Width="Auto"
    Height="Auto"
    ItemsSource="{x:Bind Chessboard}">
    <GridView.ItemsPanel>
        <ItemsPanelTemplate>
            <ItemsWrapGrid MaximumRowsOrColumns="8" Orientation="Horizontal" />