ashokpant

#!/bin/bash

# install CUDA Toolkit v9.0
# instructions from https://developer.nvidia.com/cuda-downloads (linux -> x86_64 -> Ubuntu -> 16.04 -> deb)
CUDA_REPO_PKG="cuda-repo-ubuntu1604-9-0-local_9.0.176-1_amd64-deb"
wget https://developer.nvidia.com/compute/cuda/9.0/Prod/local_installers/${CUDA_REPO_PKG}
sudo dpkg -i ${CUDA_REPO_PKG}
sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda-9-0

ahmetozlu

#----------------------------------------------
#--- Author         : Ahmet Ozlu
#--- Mail           : ahmetozlu93@gmail.com
#--- Date           : 21st September 2017
#----------------------------------------------

import face_recognition
import cv2
import os
import create_csv

ndevenish

cmake_minimum_required(VERSION 3.5)

# Find python and Boost - both are required dependencies
find_package(PythonLibs 2.7 REQUIRED)
find_package(Boost COMPONENTS python REQUIRED)

# Without this, any build libraries automatically have names "lib{x}.so"
set(CMAKE_SHARED_MODULE_PREFIX "")

# Add a shared module - modules are intended to be imported at runtime.

sonots

Usually, located at /usr/local/cuda/bin

Non-Visual Profiler

$ nvprof python train_mnist.py

I prefer to use --print-gpu-trace.

kkweon

import numpy as np
import gym
import random

from collections import deque
from keras.layers import Input, Activation, Dense, Flatten, RepeatVector, Reshape
from keras.layers.convolutional import Conv2D
from keras.models import Model
from keras import backend as K

karpathy

"""
Minimal character-level Vanilla RNN model. Written by Andrej Karpathy (@karpathy)
BSD License
"""
import numpy as np

# data I/O
data = open('input.txt', 'r').read() # should be simple plain text file
chars = list(set(data))
data_size, vocab_size = len(data), len(chars)

henrik

tmux cheatsheet

As configured in my dotfiles.

start new:

tmux

start new with session name: