delta2323

import six
import numpy

import chainer


class StatelessRNN(chainer.Chain):

    # state_names defines the order of states                                                                                                                                                                                                                                                                                              
    state_names = ('c', 'h')

delta2323

import six
import numpy

import chainer


class StatelessRNN(chainer.Chain):

    # state_names defines the order of states                                                                                                                                                                                                                                      
    state_names = ('c', 'h')

delta2323

#!/usr/bin/env python                                                                                                                                                                                              

import numpy
import six
from matplotlib import pyplot


seed = 0
N = 100
M = 10

delta2323

import time


global_hooks = {}


class FunctionHook(object):

    def __enter__(self):
        global global_hooks

delta2323

import chainer
from chainer import links
from chainer import optimizers
from chainer import functions
import numpy
import six


init = numpy.random.uniform(-1, 1, (2, 3)).astype(numpy.float32)
target = chainer.Variable(numpy.ones_like(init))

delta2323

# GPU to use: Tesla K40
# Comments are GPU memory usage measured by nvidia-smi

import chainer, cupy, numpy

cupy.cuda.Device(1).use()  
a = cupy.array((), dtype=numpy.float32)  # 74MiB
# I have not investigated what components consume first 74MiB memory.

b = cupy.ones((1024, 1024), dtype=numpy.float32)  # 78MiB

delta2323

from chainer import links
from chainer import serializers
import numpy

cls = links.Classifier(links.Linear(10, 10))
cls.add_persistent('k', numpy.random.uniform(-1, 1, (1, 3)))
print cls._persistent # => ['k']                                                                                                                                                                                   
print cls.k # => [[-0.76103399  0.09844306 -0.25231615]]                                                                                                                                                           
serializers.save_hdf5('test.hd5', cls)

delta2323

#/usr/bin/env python

import re
import requests


r = requests.get('https://papers.nips.cc/book/advances-in-neural-information-processing-systems-28-2015')
urls = re.findall('<a href="(/paper/[0-9]+.*?)">.*?</a>', r.text)
for url in urls:
    url = 'https://papers.nips.cc/' + url +'.pdf'

delta2323

#!/usr/bin/env python                                                                                                                                                                

import chainer
from chainer import functions as F
from chainer import optimizers as O
import numpy as np


print(chainer.__version__)

delta2323

# caffemodelの入手
cd models/bvlc_reference_caffenet/
wget http://dl.caffe.berkeleyvision.org/bvlc_reference_caffenet.caffemodel
# このURLはreadme.mdcaffemodel_urlを参照している
# ダウンロードに結構時間がかかる（30分くらい？）

# imagenet_mean.binaryprotoとsynset_words.txtの入手
cd ../../
./data/ilsvrc12/get_ilsvrc_aux.sh