dniku’s gists

dniku / problem_bool.py

Last active December 18, 2015 20:38

Problem Q from http://ts.lokos.net/statements/130610_ln.pdf

	to_bool = {'T': True, 'F': False}

	def matches(what, whom):
	return (whom == '?') or (what == to_bool[whom])

	def simplify(answers):
	if len(answers) > 1:
	return '?'
	return ['F', 'T'][answers.pop()]

dniku / tiled_matrix_multiplication.cu

Last active August 29, 2015 14:14

Tiled matrix multiplication with CUDA, the algorithm is from https://www.coursera.org/course/hetero

	#define TILE_WIDTH 16

	// NOTE: Since TILE_WIDTH is a constant, I cannot use the more readable idiom
	// ceil(whatever/16.0) for calculating the number of tiles. Instead I use
	// (whatever + TILE_WIDTH - 1) / TILE_WIDTH, which is equivalent. Trust me.

	// Compute C = A * B
	__global__ void matrixMultiplyShared(float A, float B, float *C, int numARows,
	int numAColumns, int numBRows,
	int numBColumns, int numCRows,

dniku / pylearn2_convnet_loader.py

Last active August 29, 2015 14:14

Loading external digit images for pylearn2 convnet example

	import cv2, numpy as np
	import theano
	from pylearn2.utils import serial

	def load_pl2_recognizer(model_path):
	model = serial.load(model_path)

	X = model.get_input_space().make_theano_batch()
	Y = model.fprop(X)

dniku / setup_pylearn2_aws.sh

Last active August 29, 2015 14:15

To launch, do "chmod +x setup_pylearn2_aws.sh" first. An updated version of http://www.kurtsp.com/deep-learning-in-python-with-pylearn2-and-amazon-ec2.html

	sudo apt-get -y update
	sudo apt-get -y upgrade
	sudo apt-get -y dist-upgrade
	sudo apt-get -y install git make python-dev python-setuptools libblas-dev gfortran g++ python-pip python-numpy python-scipy liblapack-dev
	sudo pip install ipython nose
	sudo apt-get install screen
	sudo pip install --upgrade git+git://github.com/Theano/Theano.git
	wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_6.5-14_amd64.deb
	sudo dpkg -i cuda-repo-ubuntu1404_6.5-14_amd64.deb
	sudo apt-get update

dniku / amazon_cuda_log.txt

Last active August 29, 2015 14:15

Shell output on an Amazon g2.2xlarge instance for: wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_6.5-14_amd64.deb && sudo dpkg -i cuda-repo-ubuntu1404_6.5-14_amd64.deb && sudo apt-get update && sudo apt-get install cuda

	--2015-02-09 16:39:38-- http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_6.5-14_amd64.deb
	Resolving developer.download.nvidia.com (developer.download.nvidia.com)... 92.123.72.158, 92.123.72.176
	Connecting to developer.download.nvidia.com (developer.download.nvidia.com)\|92.123.72.158\|:80... connected.
	HTTP request sent, awaiting response... 200 OK
	Length: 2138 (2.1K) [application/x-deb]
	Saving to: ‘cuda-repo-ubuntu1404_6.5-14_amd64.deb’

	100%[==================================================================================================>] 2,138 --.-K/s in 0s

	2015-02-09 16:39:39 (406 MB/s) - ‘cuda-repo-ubuntu1404_6.5-14_amd64.deb’ saved [2138/2138]

dniku / aws_pylearn2_mnist_maxout

Created February 9, 2015 19:33

	ubuntu@ip-172-31-44-36:~/pylearn2/pylearn2/scripts/papers/maxout$ THEANO_FLAGS="device=gpu,floatX=float32" ~/pylearn2/pylearn2/scripts/train.py mnist.yaml
	Using gpu device 0: GRID K520

	(train.py:4883): Gdk-CRITICAL **: gdk_cursor_new_for_display: assertion 'GDK_IS_DISPLAY (display)' failed
	Traceback (most recent call last):
	File "/home/ubuntu/pylearn2/pylearn2/scripts/train.py", line 252, in <module>
	args.verbose_logging, args.debug)
	File "/home/ubuntu/pylearn2/pylearn2/scripts/train.py", line 197, in train
	train_obj = serial.load_train_file(config)
	File "/home/ubuntu/pylearn2/pylearn2/utils/serial.py", line 524, in load_train_file

dniku / Still running MNIST maxout

Created February 12, 2015 16:13

	Using gpu device 0: GRID K520
	error: XDG_RUNTIME_DIR not set in the environment.
	error: XDG_RUNTIME_DIR not set in the environment.

	(train.py:2434): Gdk-CRITICAL **: gdk_cursor_new_for_display: assertion 'GDK_IS_DISPLAY (display)' failed
	/home/ubuntu/pylearn2/pylearn2/utils/image.py:16: UserWarning: Unable to import matplotlib. Some features unavailable. Original exception: constructor returned NULL
	"Original exception: " + str(matplotlib_exception))

	(train.py:2434): Gdk-CRITICAL **: gdk_cursor_new_for_display: assertion 'GDK_IS_DISPLAY (display)' failed
	/home/ubuntu/pylearn2/pylearn2/sandbox/cuda_convnet/__init__.py:66: UserWarning: You are using probably a too old Theano version. That will cause compilation crash. If so, update Theano.

dniku / blank_template.html

Created February 15, 2015 21:54

	<!doctype html>
	<html>
	<head>
	<meta http-equiv="content-type" content="text/html; charset=utf-8">
	<title>Sandbox</title>
	</head>
	<body>

	<script src="sandbox.js"></script>
	</body>

dniku / expansions.py

Created March 19, 2015 17:54

Expansions of a natural number into sums

	def get_expansions(num, nterms):
	result = []

	def run(cur_sum, cur_vals, remain):
	if remain == 0:
	if cur_sum == num:
	result.append(tuple(cur_vals))
	return
	elif remain == 1:
	cur_vals.append(num - cur_sum)

dniku / classify_caffe.py

Last active August 29, 2015 14:17

	from __future__ import division, print_function

	from pprint import pprint
	import glob
	import caffe

	MODEL_FILE = '../models/lenet.prototxt'
	PRETRAINED_FILE = '../models/lenet_pretrained.caffemodel'
	IMAGE_DIR = 'test'

Dmitry dniku