bwasti

import torch

def cudagraph(f):
    _graphs = {}
    def f_(*args):
        key = hash(tuple(tuple(a.shape) for a in args))
        if key in _graphs:
            wrapped, *_ = _graphs[key]
            return wrapped(*args)
        g = torch.cuda.CUDAGraph()

bwasti

console.log("Oh no! bun hit an error\n")
const report = 'bun report blah blah blah'
const url = `https://duckduckgo.com/?q=${report}`
const str = '[click to upload report]'
console.log(`\u001b]8;;${url}\u001b\\${str}\u001b]8;;\u001b\\`)

bwasti

read -p "this script will remove libarrayfire from your system and install all requirements to build from source. continue? [Y/n]" -n 1 -r
echo
if [[ $REPLY =~ ^[Yy]$ ]]
then
  sudo apt remove libarrayfire-dev libarrayfire-cpu3 libarrayfire-cpu-dev
  sudo apt install -y libblas-dev liblapack-dev liblapacke-dev libfftw3-dev libboost-all-dev cmake make g++
  cd /tmp
  rm -rf arrayfire
  git clone https://github.com/arrayfire/arrayfire.git
  cd arrayfire

bwasti

# conv bwd implemented with fwd functions

import torch
import torch.nn.functional as F


def dconv2d(grad, x, w, stride, padding, groups):
    batch = grad.shape[0]
    channel_out = grad.shape[1]
    channel_in = x.shape[1]

bwasti

# example of backward pass implemented with only forward functions

import torch
import torch.nn.functional as F


def dconv(grad, x, w, stride, padding, groups):
    batch = grad.shape[0]
    channel_out = grad.shape[1]
    channel_in = x.shape[1]

bwasti

# examples of backward passes implemented with fwd functions

import torch
import torch.nn.functional as F


def simple():
    print("simple")
    x = torch.randn(1, 1, 4, 4)
    x.requires_grad = True

bwasti

Shumai's io namespace

Inference

Want to expose your machine learning model to the network? Use io.serve:

import * as sm from '@shumai/shumai'
import { model } from './model'

bwasti

import os

for cmd in ["add", "mm"]:
    print(cmd)
    for n in range(0, 14):
        N = 2**n
        best_cpu = 0
        best_cuda = 0
        for thread in [1, 2, 4, 8, 16, 32, 64, 80]:
            full_cmd = f"OMP_NUM_THREADS={thread} python comp.py {cmd} {N}"

bwasti

import torch
import time
import sys

fn = sys.argv[1]
N = int(sys.argv[2])
iters = 1000
mps = torch.device("mps")

a = torch.randn(N, N)

bwasti

#include <cstdint>

extern "C" {

void init() {}

int64_t bytesUsed() {
  return 0;
}