Ulisse Mini UlisseMini

## sympy-exact-approx.py
# Run with ipython -i sympy-exact-approx.py so you can play with it afterwards

import sympy as sp

x = sp.Symbol('x')

# function to approximate, if this is exotic change inner() to compute
# integrals numerically instead of symbolically.
def f(x):
    # standard example

## flatten-hessian-mnist.py
# fuller code https://colab.research.google.com/drive/12zXLbykv537MrZr6WDCnRIqKQ5h8UjVw?usp=sharing

fn = lambda *params: F.nll_loss(stateless.functional_call(model, {n: p for n,p in zip(names, params)}, x), y)
H = hessian(fn, tuple(model.parameters()))
# H[i][j] contains the derivatives of the loss with respect to every parameter in model.parameters()[i] and [j].
# (It's an annoying tuple)

# flatten the annoying tuple!
rows = []
shapes = [p.shape for p in model.parameters()]

## bunny.py
N = 100

def guess(P, j):
  # after setting P[j] = 0 we have removed P[j] probability mass from
  # the distribution, so the new sum is 1-P[j]. Dividing everything by
  # 1-P[j] fixes this. (a minor detail, instead of 1 we use sum(P) for
  # numerical stability since sum(P) isn't exactly 1)
  s = sum(P) - P[j]
  return [
    0 if i == j else P[i] / s

## torus.py
import matplotlib.pyplot as plt
import numpy as np
tau = 2*np.pi

fig = plt.figure()
ax = fig.add_subplot(projection='3d')

# map from unit square to surface of a torus
def torus(u, v):
    X = (2 + np.cos(tau*v))*np.cos(tau*u)

## plot.py
import plotly.graph_objects as go
import plotly.express as px
import numpy as np
import pandas as pd

# constants
m, n = 6, 5
pi = np.pi

def vec_x(theta, phi):

## calc-chess-games.py
import chess
import statistics
import random

def median_move_count(b, depth = 3):
    if depth < 0: return []

    mvs = []
    mvs.append(len(list(b.legal_moves)))
    for move in b.legal_moves:

## owo.html
<!DOCTYPE html>
<html>
  <head>
    <meta charset="utf-8">
    <title>owo</title>

    <style>
      html {
        background: #222;
        color: #fff;

## run.js
const Docker = require('dockerode')
const docker = new Docker()
const stripAnsi = require("strip-ansi");

// docker run --cpus=0.1 --memory=64m --kernel-memory=64m --ulimit="nproc=128:128" -i alpine:latest sh -c "$*"

const cmd = process.argv.join(' ')

const kb = 1024
const mb = kb * 1024

## blind-regex-based-nosqli.py
"""
Inspired from https://youtu.be/NO_lsfhQK_s (aka blind regex based nosql injection)
"""

import re, time, string

should_sleep = True
binary_search_blind = True
attempts = 0

## asm.py
import time
import struct

DEBUG = True

class Instruction():
    def __init__(self, opcode, args, name):
        self.opcode = opcode
        self.args = args
        self.name = name
	# Run with ipython -i sympy-exact-approx.py so you can play with it afterwards

	import sympy as sp

	x = sp.Symbol('x')

	# function to approximate, if this is exotic change inner() to compute
	# integrals numerically instead of symbolically.
	def f(x):
	# standard example
	# fuller code https://colab.research.google.com/drive/12zXLbykv537MrZr6WDCnRIqKQ5h8UjVw?usp=sharing

	fn = lambda *params: F.nll_loss(stateless.functional_call(model, {n: p for n,p in zip(names, params)}, x), y)
	H = hessian(fn, tuple(model.parameters()))
	# H[i][j] contains the derivatives of the loss with respect to every parameter in model.parameters()[i] and [j].
	# (It's an annoying tuple)

	# flatten the annoying tuple!
	rows = []
	shapes = [p.shape for p in model.parameters()]
	N = 100

	def guess(P, j):
	# after setting P[j] = 0 we have removed P[j] probability mass from
	# the distribution, so the new sum is 1-P[j]. Dividing everything by
	# 1-P[j] fixes this. (a minor detail, instead of 1 we use sum(P) for
	# numerical stability since sum(P) isn't exactly 1)
	s = sum(P) - P[j]
	return [
	0 if i == j else P[i] / s
	import matplotlib.pyplot as plt
	import numpy as np
	tau = 2*np.pi

	fig = plt.figure()
	ax = fig.add_subplot(projection='3d')

	# map from unit square to surface of a torus
	def torus(u, v):
	X = (2 + np.cos(tauv))np.cos(tau*u)
	import plotly.graph_objects as go
	import plotly.express as px
	import numpy as np
	import pandas as pd

	# constants
	m, n = 6, 5
	pi = np.pi

	def vec_x(theta, phi):
	import chess
	import statistics
	import random

	def median_move_count(b, depth = 3):
	if depth < 0: return []

	mvs = []
	mvs.append(len(list(b.legal_moves)))
	for move in b.legal_moves:
	<!DOCTYPE html>
	<html>
	<head>
	<meta charset="utf-8">
	<title>owo</title>

	<style>
	html {
	background: #222;
	color: #fff;
	const Docker = require('dockerode')
	const docker = new Docker()
	const stripAnsi = require("strip-ansi");

	// docker run --cpus=0.1 --memory=64m --kernel-memory=64m --ulimit="nproc=128:128" -i alpine:latest sh -c "$*"

	const cmd = process.argv.join(' ')

	const kb = 1024
	const mb = kb * 1024
	"""
	Inspired from https://youtu.be/NO_lsfhQK_s (aka blind regex based nosql injection)
	"""

	import re, time, string

	should_sleep = True
	binary_search_blind = True
	attempts = 0
	import time
	import struct

	DEBUG = True

	class Instruction():
	def __init__(self, opcode, args, name):
	self.opcode = opcode
	self.args = args
	self.name = name