fancellu

ComfyUI Desktop (0.43.0) — QwenVL GGUF + Attention Backends on Windows 11

Environment: RTX 4090, Python 3.12 (cp312), torch 2.9.1+cu130, triton-windows 3.2.0.post21

1. QwenVL GGUF (llama-cpp-python fork)

The standard llama-cpp-python on PyPI doesn't include the vision chat handlers. You need JamePeng's fork as a pre-built wheel.

My python was in D:\comfyui_run\ComfyUI\.venv\Scripts\python.exe

fancellu

<!DOCTYPE html>
<body style="font-family:sans-serif; padding:30px; background:#1e1e1e; color:#eee;">
    <h2>RSA Binary Verifier (No Size Limit)</h2>
    
    <label>1. Public Key (PEM):</label><br>
    <textarea id="p" style="width:100%; height:120px; background:#333; color:#fff; border:1px solid #555;"></textarea><br><br>
    
    <label>2. Signature (Base64):</label><br>
    <input id="s" type="text" style="width:100%; background:#333; color:#fff; border:1px solid #555;"><br><br>
    

fancellu

ZIO Beam WordCount (with added Cats Effect now!)

A simple word count program built with ZIO and Apache Beam/Scio that demonstrates retry logic and error handling.

Features

• Counts words in a text file using Apache Beam
• Built with ZIO for functional effect management or Cats Effect and cats-retry
• Includes retry logic with fixed delays
• Simulates random failures for testing resilience

fancellu

2 files:

batch-1.xml

<?xml version="1.0" encoding="UTF-8"?><root><order id="ord-555" region="UK">
        <line product-id="p1" qty="3"/>        
    </order><order id="ord-666" region="MARS">
        <line product-id="p1" qty="1"/>
        <line product-id="p2" qty="2"/>
        <line product-id="p99" qty="1"/> 

fancellu

{
  "items": [
    { "product":"Laptop", "qty":1, "line_total":1200 },
    { "product":"Mouse", "qty":2, "line_total":100 },
    { "error":"Product not found: p99", "product":"UNKNOWN", "qty":1, "line_total":0 }
  ],
  "total_cost": 1300,
  "order_id": "ord-666",
  "tax_rate": 0
}

fancellu

import pandas as pd
import plotly.graph_objects as go
import gradio as gr
from io import StringIO


def create_interactive_bitcoin_chart(csv_text=None):
    """Create interactive Bitcoin volume chart from CSV data"""
    try:
        if csv_text and csv_text.strip():

fancellu

import gradio as gr
import numpy as np
import spaces
import torch
import random
from PIL import Image

from diffusers import FluxKontextPipeline
from diffusers.utils import load_image

fancellu

https://huggingface.co/spaces/Fancellu/Kokoro-TTS-Zero-CPU

To run from docker

docker run -it -p 7860:7860 --platform=linux/amd64 registry.hf.space/fancellu-kokoro-tts-zero-cpu:latest python app.py

fancellu

https://huggingface.co/spaces/Fancellu/bark-pytorch2.6-compatible

This solves the PyTorch 2.6 breaking change where weights_only=True became the default, causing issues with older model files that contain non-tensor objects, with monkey patch patched_load()

Updated to Python 3.11, latest Gradio, and requirements.txt

If you have a GPU, you can run from Docker Desktop directly

docker run -it -p 7860:7860 --gpus=all --platform=linux/amd64 registry.hf.space/fancellu-bark-pytorch2-6-compatible:latest python app.py

fancellu

import zio._
import scala.{ Console => SConsole }
import scala.io.StdIn
import java.io.{ BufferedReader, IOException }
import scala.util.Try

object InterruptableReadLine extends ZIOAppDefault {

  def altReadLine(reader: BufferedReader = SConsole.in) =
    ZIO