bartvm

import array
import collections
import bisect

class MKAverage:

    def __init__(self, m: int, k: int):
        # self.q = collections.deque()
        self.q = array.array("L")
        self.p = None

bartvm

class MKAverage:

    def __init__(self, m: int, k: int):
        self.q = SkipList(6, 0.5)
        self.m = m
        self.k = k
        
        self.counter = 0
        
        self.sum_ = 0

bartvm

import math
import random
  
class Node:
    __slots__ = ["key", "forward", "backward", "next"]
    
    def __init__(self, key, level):
        self.key = key
        self.forward = [None] * (level + 1)
        self.backward = [None] * (level + 1)

bartvm

Anonymous ballot

The goal is to allow a user to anonymously share sensitive information which can be used in a selection process. The user must be able to prove they were admitted without revealing their sensitive information. A two-step ring signature process meets these requirements.

Ring signature

Consider a group of n users where each user i has a public, P_i, and secret (private) key, S_i. A ring signature allows a user to sign a message using their private key, s = sign(m, S_i, P₁, ..., P_n), proving that they are a member of the group without revealing their identity.

Approach

bartvm

def f(x, y):
    z = x * y
    return z

# Operator overloading
adjoints = {
    'mul': lambda dz, x, y: dz * y, dz * x
}

def f(x, y):

bartvm

diff --git a/torch/csrc/autograd/engine.cpp b/torch/csrc/autograd/engine.cpp
index f1e09b0e..c50f507e 100644
--- a/torch/csrc/autograd/engine.cpp
+++ b/torch/csrc/autograd/engine.cpp
@@ -21,6 +21,10 @@
 #include <THC/THC.h>
 #endif
 
+void tid() {
+    printf("%d ", (int)std::hash<std::thread::id>()(std::this_thread::get_id()));

bartvm

diff --git a/torch/csrc/autograd/engine.cpp b/torch/csrc/autograd/engine.cpp
index f1e09b0e..d78c03e8 100644
--- a/torch/csrc/autograd/engine.cpp
+++ b/torch/csrc/autograd/engine.cpp
@@ -110,7 +110,7 @@ Engine::~Engine() = default;
 auto Engine::thread_main(std::shared_ptr<ReadyQueue> queue, int device) -> void {
   THInferNumThreads();
   AutoGPU guard(device);
-  while (1) {
+  while (!exit.back().load()) {

bartvm

from torch import Tensor
from torch.autograd import Function, Variable

class Foo(Function):
  def forward(self, x):
    return x

  def backward(self, dz_star):
    dphidz = Variable(Tensor(1), requires_grad=True)
    dphidz.backward()

bartvm

import ast
import collections
import inspect
import numbers
import textwrap

import numpy

PUSH = ast.Attribute(value=ast.Name(id='_stack', ctx=ast.Load()),
                     attr='push', ctx=ast.Load())

bartvm

class Linear:
	def __init__(self, in_dim, out_dim):
		self.params = {
			'W': np.random.rand(in_dim, out_dim)
			'b': np.random.rand(out_dim)
		}

	def run(self, x):
		return sigmoid(self.params['W'] @ x + self.params['b'])