def wrapped(self, asin, is_cache=True):
from amazon import Amazon
md5 = hashlib.md5(asin)
key = '%s.gz' % md5.hexdigest()
product_info = Amazon.load_cache(key)
if 'modified_date' in product_info and (
datetime.datetime.now() - product_info['modified_date']
).days < Amazon.CACHE_EXPIRED_DAYS:
pass
else:
self.load('https://www.amazon.com/dp/%s' % asin)
page = self.lr.body
image_urls = []
for ele in self.lr.xpaths('//div[@class="imgTagWrapper"]/img'):
if 'data-old-hires' in ele.attrib:
image_urls.append(ele.attrib['data-old-hires'])
product_info['page'] = page
product_info['image_urls'] = image_urls
product_info['modified_date'] = datetime.datetime.now()
Amazon.save_cache(key, product_info)
return detail(self,
asin,
is_cache=is_cache,
page_info=product_info)
def amazon_info(request):
title = request.params["title"]
author = request.params["author"]
keys = AmazonKeys()
amazon = Amazon(keys.access_key, keys.secret_access_key, keys.associate_tag)
amazon.setProxy("proxy.ome.toshiba.co.jp")
result = {}
try:
result_xml = amazon.itemSearch("Books", Title=u"獄門島", Author=u"横溝正史", ResponseGroup="Medium") # 本
dom = xml.dom.minidom.parseString(result_xml)
items = dom.getElementsByTagName("Item")
if len(items) != 0:
item = items[0]
result["image"] = getValue(item, "MediumImage", "URL")
result["url"] = getValue(item, "DetailPageURL")
result["title"] = title
print getValue(item, "ASIN"), getValue(item, "LargeImage", "URL"), getValue(
item, "DetailPageURL"
), getValue(item, "Author"), getValue(item, "Title")
dom.unlink()
except HTTPError, e:
print e.code
def get_asin(request):
output = {}
if request.method == "GET":
if request.GET.get("url"):
amazon = Amazon()
url = request.GET.get("url")
asin = amazon.get_asin_from_url(url)
output["asin"] = asin
return HttpResponse(json.dumps(output), content_type="application/json")
def save(self):
if not self.id:
a = Amazon()
a.get_document(self.ISBN)
self.title = a.title
self.author = a.author
self.summary = a.summary
self.cover_image = a.cover_image
self.book_image = settings.MEDIA_URL + '/media/book_' + str(random.randint(1,9)) + '.png'
self.asin = a.asin
super(Book, self).save()
def checkout_for_list(request, list):
lists = get_object_or_404(List, id=list)
output = {}
amazon = Amazon()
amazon.get_items(list=list)
cart = amazon.get_cart()
output["url"] = "%s" % (cart.Cart.PurchaseURL)
output["subtotal"] = "%s" % (cart.Cart.SubTotal.FormattedPrice)
print cart.Cart.PurchaseURL
return HttpResponse(json.dumps(output), content_type="application/json")
def get_item_for_asin(request, asin):
output = {}
if request.method == "GET":
amazon = Amazon()
item = amazon.get_item_by_asin(asin)
output["asin"] = str(asin)
output["url"] = str(item.Items.Item.DetailPageURL)
output["manufacturer"] = str(item.Items.Item.ItemAttributes.Manufacturer)
output["product_group"] = str(item.Items.Item.ItemAttributes.ProductGroup)
output["title"] = str(item.Items.Item.ItemAttributes.Title)
output["description"] = str(item.Items.Item.ItemLinks.ItemLink.Description)
output["short_url"] = "http://amzn.com/%s" % asin
return HttpResponse(json.dumps(output), content_type="application/json")
def con_search(keyword):
amazon = Amazon("", "")
xml = amazon.itemSearch("Books", Keywords=keyword, ItemPage="1")
# XMLから情報を取り出す
from BeautifulSoup import BeautifulStoneSoup
soup = BeautifulStoneSoup(xml)
#10個のコンテンツをコンテンツテーブルへ格納
for item in soup.findAll('item'):
isbn10 = item.asin.contents[0]
#issbn13
#isbn13 = item.ean.contents[0]
isbn13 = ''
#author
try: #author（著者）はない場合があるのでエラー処理
author = item.author.contents[0]
except AttributeError:
author = ""
else:
author = item.author.contents[0]
title = item.title.contents[0]
#url
url = item.detailpageurl.contents[0]
#image_url
try:
image_url = item.largeimage.url.contents[0]
except NameError:
image_url = ""
except AttributeError:
image_url = ""
else:
image_url = item.largeimage.url.contents[0]
arasuji = HTMLParser(url)
#文字参照を変換する
arasuji = htmlentity2utf8(arasuji)
#あらすじからいらない文字、記号を取り除く
#print arasuji
con_tag = extractKeyword(arasuji)
con_tag = ','.join(con_tag)
#print title,isbn10
#DBへ格納する
dbupdate(isbn10, isbn13, title, author, image_url, arasuji, con_tag)
return
def getAmazon(isbn):
result = []
amazon = Amazon(access_key, access_secret, associate_tag)
xml = amazon.itemLookup(isbn,
SearchIndex='Books',
IdType='ISBN',
ReviewPage='1',
ResponseGroup='Reviews')
soup = BeautifulSoup(xml)
if soup.find('hasreviews').getText() == 'false':
return result
url = soup.find('iframeurl').contents[0]
iframe = urllib.urlopen(url.replace('&', '&'))
soup = BeautifulSoup(iframe)
for div in soup.findAll('div', {'style': 'margin-left:0.5em;'}):
for i in div.findAll('div'):
i.replaceWith('')
result.append(div.getText())
return result
print(f'wall: {time.time() - start}')
f.write(f'wall: {time.time() - start}')
if __name__ == '__main__':
args = config()
devices = list(map(int, args.gpu.split(',')))
n_gpus = len(devices)
# For GCMC based on sample, we require node has its own features.
# Otherwise (node_id is the feature), the model can not scale
dataset = Amazon(args.data_name,
'cpu',
mix_cpu_gpu=args.mix_cpu_gpu,
use_one_hot_fea=args.use_one_hot_fea,
symm=args.gcn_agg_norm_symm,
test_ratio=args.data_test_ratio,
valid_ratio=args.data_valid_ratio)
print("Loading data finished ...\n")
args.src_in_units = dataset.user_feature_shape[1]
args.dst_in_units = dataset.item_feature_shape[1]
args.rating_vals = dataset.possible_rating_values
# cpu
if devices[0] == -1:
run(0, 0, args, ['cpu'], dataset)
# gpu
elif n_gpus == 1:
run(0, n_gpus, args, devices, dataset)
from amazon import Amazon
from pesquisa import Pesquisa
pesquisa = Pesquisa()
produto = 'iphone'
conteudo = pesquisa.procurarProduto(produto)
amazon = Amazon()
amazon.recuperarProdutos(conteudo)
amazon.recuperarPrecos(conteudo)
amazon.gravarDados('produto2')
pesquisa.finalizar()
from database import DataBase
from config import Config
from amazon import Amazon
if __name__ == '__main__':
config = Config()
dbData = config.data['database']
awsData = config.data['aws']
# Connect to MySQL DB
db = DataBase(dbData['host'], dbData['user'], dbData['pass'], dbData['name'])
# Connect to AWS
amazon = Amazon(awsData['key_name'], awsData['secret_key'])
if hasattr(config, 'domains') and config.domains == ['all']:
config.domains = []
domains_list = db.get_all(db.cursor, 'domains')
if domains_list == []:
sys.exit('No domains records on database.')
for d in domains_list:
config.domains.append(d['name'])
if config.action == 'check':
for domain_name in config.domains:
domain = db.get(db.cursor, 'domains', 'name', domain_name)
if not domain:
print "[ERROR] The domain doesn't exist"
sys.exit(0)
elif options.loglevel == "INFO":
logLevelInt = 3
elif options.loglevel == "WARN":
logLevelInt = 4
else:
LOG.warning("Unknown Log level %s given. Defaulting to INFO." %
options.loglevel)
widevineAdapter.SetLogLevel(logLevelInt)
if options.deleteDumpDir:
os.system("rm {0}/*".format(constants.DUMP_DIRECTORY))
if not os.path.isdir(constants.DUMP_DIRECTORY):
os.mkdir(constants.DUMP_DIRECTORY)
az = Amazon()
if URL_REGEX.match(amazonUrlOrAsin):
LOG.info("Using amazon url: %s" % amazonUrlOrAsin)
# login is needed else e.g. no asin can be retrieved
az.logIn()
titleAmazonMovieEntryMap = az.getVideoPageDetailsMapping(
amazonUrlOrAsin)
for key, val in titleAmazonMovieEntryMap.items():
LOG.info("%s: %s" % (key, val))
LOG.info("")
if {} == titleAmazonMovieEntryMap:
LOG.warning("No video data could be found! Mapping was empty.")
else:
LOG.info("Launching rendering script for ASIN %s" % amazonUrlOrAsin)
def main(secondsToAudit, configFile, gracePeriod, historyFile, logFile, auditPayments, auditRefunds):
global _config
global _civiDB
global _awsLink
global _stompLink
global _logFile
startTime = datetime.fromtimestamp(int(time.time()) - int(secondsToAudit), pytz.utc)
endTime = datetime.fromtimestamp(int(time.time()) - int(gracePeriod), pytz.utc)
print("AWS audit requested from %s to %s" % (startTime.isoformat(), endTime.isoformat()))
# === Initialize the configuration file ===
localdir = os.path.dirname(os.path.abspath(__file__))
_config = SafeConfigParser()
fileList = ["%s/amazon-audit.cfg" % localdir]
if configFile is not None:
fileList.append(configFile)
_config.read(fileList)
# === Open up ze STOMP ===
_stompLink = DistStomp(_config.get('Stomp', 'server'), _config.getint('Stomp', 'port'))
_stompLink.connect()
# === Connection to Amazon ===
_awsLink = Amazon(
awsEndpoint=_config.get('AwsConfig', 'endpoint'),
awsAccessKey=_config.get('AwsConfig', 'accessKey'),
awsSecret=_config.get('AwsConfig', 'secretKey')
)
# === Connection to MySQL ===
_civiDB = MySQL.connect(
_config.get('MySQL', 'host'),
_config.get('MySQL', 'user'),
_config.get('MySQL', 'password'),
_config.get('MySQL', 'schema')
)
# === Open up the history and log files ===
# If the history file exists, it will modify the start time of this script to be the end time of the
# history file.
hfile = None
historyStart = startTime
if historyFile and os.path.exists(historyFile):
hfile = open(historyFile, 'r')
if hfile.readline().strip() == AWS_HISTORY_FILE_VERSTR:
historyStart = dateutil.parser.parse(hfile.readline().strip())
historyEnd = dateutil.parser.parse(hfile.readline().strip())
startTime = historyEnd
print("History file modified search period, now %s to %s" % (startTime.isoformat(), endTime.isoformat()))
else:
print('Not starting with a valid history file.')
if logFile:
_logFile = open(logFile, 'a')
_logFile.write("!!! Starting run for dates %s -> %s\n" % (startTime.isoformat(), endTime.isoformat()))
# === Sanity checks ===
if endTime < startTime:
startTime = endTime
# === Main Application ===
awsTransactions = []
# --- Process all previously pending transactions from the history file. If the transaction is still in some form
# of pending, add it back to the history list.
historyCount = 0
historyList = []
historyStats = {
'Success': 0,
'Pending': 0,
'Failed': 0,
'Ignored': 0
}
if hfile:
print("Processing history file")
for txn in hfile:
historyCount += 1
awsTransactions.append(json.loads(txn))
hfile.close()
# --- Obtain AWS history ---
if auditPayments:
print("Obtaining AWS payment transactions for the period %s -> %s" % (startTime.isoformat(), endTime.isoformat()))
awsTransactions += _awsLink.getAccountActivity(startTime, endDate=endTime, fpsOperation='Pay')
print("Obtained %d transactions" % len(awsTransactions))
if auditRefunds:
print("Obtaining AWS refund transactions for the period %s -> %s" % (startTime.isoformat(), endTime.isoformat()))
awsTransactions += _awsLink.getAccountActivity(startTime, endDate=endTime, fpsOperation='Refund')
print("Obtained %d transactions" % len(awsTransactions))
# --- Main loop: checks each aws transaction against the Civi database; adding it if it doesn't exist ---
txncount = 0
for txn in awsTransactions:
txncount += 1
result = dispatchTransaction(txn, auditPayments, auditRefunds)
historyStats[result] += 1
if result == 'Pending':
historyList.append(txn)
print("\n--- Finished processing of messages. ---\n")
# --- Write the history file ---
if historyFile:
print("Rewriting history file with %d transactions" % len(historyList))
hfile = open(historyFile, 'w')
hfile.write("%s\n%s\n%s\n" % (AWS_HISTORY_FILE_VERSTR, historyStart.isoformat(), endTime.isoformat()))
for txn in historyList:
hfile.write("%s\n" % json.dumps(txn))
print("Flushing history file in preparation for main loop")
hfile.flush()
# --- Final statistics
print("Processed %d AWS messages" % txncount)
print(" ... of which %d messages were from history" % historyCount)
print("This resulted in the following:")
for entry in historyStats.items():
print(" %s Messages: %d" % entry)
# === Final Application Cleanup ===
print("\nCleaning up.")
_civiDB.close()
_stompLink.disconnect()
if hfile:
hfile.close()
if _logFile:
_logFile.close()
time.sleep(1) # Let the STOMP library catch up
def train(args):
print(args)
if args.data_name == 'electronic':
dataset = Amazon(args.data_name,
args.device,
use_one_hot_fea=args.use_one_hot_fea,
symm=args.gcn_agg_norm_symm,
test_ratio=args.data_test_ratio,
valid_ratio=args.data_valid_ratio)
else:
dataset = MovieLens(args.data_name,
args.device,
use_one_hot_fea=args.use_one_hot_fea,
symm=args.gcn_agg_norm_symm,
test_ratio=args.data_test_ratio,
valid_ratio=args.data_valid_ratio)
print("Loading data finished ...\n")
args.src_in_units = dataset.user_feature_shape[1]
if args.data_name == 'electronic':
args.dst_in_units = dataset.item_feature_shape[1]
else:
args.dst_in_units = dataset.item_feature_shape[1]
args.rating_vals = dataset.possible_rating_values
start = time.time()
### build the net
net = Net(args=args)
net = net.to(args.device)
nd_possible_rating_values = th.FloatTensor(
dataset.possible_rating_values).to(args.device)
rating_loss_net = nn.CrossEntropyLoss()
learning_rate = args.train_lr
optimizer = get_optimizer(args.train_optimizer)(net.parameters(),
lr=learning_rate)
print("Loading network finished ...\n")
### perpare training data
train_gt_labels = dataset.train_labels
train_gt_ratings = dataset.train_truths
### prepare the logger
train_loss_logger = MetricLogger(
['iter', 'loss', 'rmse'], ['%d', '%.4f', '%.4f'],
os.path.join(args.save_dir, 'train_loss%d.csv' % args.save_id))
valid_loss_logger = MetricLogger(['iter', 'rmse'], ['%d', '%.4f'],
os.path.join(
args.save_dir,
'valid_loss%d.csv' % args.save_id))
test_loss_logger = MetricLogger(['iter', 'rmse'], ['%d', '%.4f'],
os.path.join(
args.save_dir,
'test_loss%d.csv' % args.save_id))
### declare the loss information
best_valid_rmse = np.inf
no_better_valid = 0
best_iter = -1
count_rmse = 0
count_num = 0
count_loss = 0
dataset.train_enc_graph = dataset.train_enc_graph.int().to(args.device)
dataset.train_dec_graph = dataset.train_dec_graph.int().to(args.device)
dataset.valid_enc_graph = dataset.train_enc_graph
dataset.valid_dec_graph = dataset.valid_dec_graph.int().to(args.device)
dataset.test_enc_graph = dataset.test_enc_graph.int().to(args.device)
dataset.test_dec_graph = dataset.test_dec_graph.int().to(args.device)
print("Start training ...")
dur = []
for iter_idx in range(1, args.train_max_iter):
if iter_idx > 3:
t0 = time.time()
net.train()
if args.data_name == 'electronic':
pred_ratings = net(dataset.train_enc_graph,
dataset.train_dec_graph, dataset.user_feature,
dataset.item_feature)
else:
pred_ratings = net(dataset.train_enc_graph,
dataset.train_dec_graph, dataset.user_feature,
dataset.item_feature)
loss = rating_loss_net(pred_ratings, train_gt_labels).mean()
count_loss += loss.item()
optimizer.zero_grad()
loss.backward()
nn.utils.clip_grad_norm_(net.parameters(), args.train_grad_clip)
optimizer.step()
if iter_idx > 3:
dur.append(time.time() - t0)
if iter_idx == 1:
print("Total #Param of net: %d" % (torch_total_param_num(net)))
print(
torch_net_info(net,
save_path=os.path.join(
args.save_dir, 'net%d.txt' % args.save_id)))
real_pred_ratings = (th.softmax(pred_ratings, dim=1) *
nd_possible_rating_values.view(1, -1)).sum(dim=1)
rmse = ((real_pred_ratings - train_gt_ratings)**2).sum()
count_rmse += rmse.item()
count_num += pred_ratings.shape[0]
if iter_idx % args.train_log_interval == 0:
train_loss_logger.log(iter=iter_idx,
loss=count_loss / (iter_idx + 1),
rmse=count_rmse / count_num)
logging_str = "Iter={}, loss={:.4f}, rmse={:.4f}, time={:.4f}".format(
iter_idx, count_loss / iter_idx, count_rmse / count_num,
np.average(dur))
count_rmse = 0
count_num = 0
if iter_idx % args.train_valid_interval == 0:
valid_rmse = evaluate(args=args,
net=net,
dataset=dataset,
segment='valid')
valid_loss_logger.log(iter=iter_idx, rmse=valid_rmse)
logging_str += ',\tVal RMSE={:.4f}'.format(valid_rmse)
if valid_rmse < best_valid_rmse:
best_valid_rmse = valid_rmse
no_better_valid = 0
best_iter = iter_idx
test_rmse = evaluate(args=args,
net=net,
dataset=dataset,
segment='test')
best_test_rmse = test_rmse
test_loss_logger.log(iter=iter_idx, rmse=test_rmse)
logging_str += ', Test RMSE={:.4f}'.format(test_rmse)
else:
no_better_valid += 1
if no_better_valid > args.train_early_stopping_patience\
and learning_rate <= args.train_min_lr:
logging.info(
"Early stopping threshold reached. Stop training.")
break
if no_better_valid > args.train_decay_patience:
new_lr = max(learning_rate * args.train_lr_decay_factor,
args.train_min_lr)
if new_lr < learning_rate:
learning_rate = new_lr
logging.info("\tChange the LR to %g" % new_lr)
for p in optimizer.param_groups:
p['lr'] = learning_rate
no_better_valid = 0
if iter_idx % args.train_log_interval == 0:
print(logging_str)
print('Best Iter Idx={}, Best Valid RMSE={:.4f}, Best Test RMSE={:.4f}'.
format(best_iter, best_valid_rmse, best_test_rmse))
train_loss_logger.close()
valid_loss_logger.close()
test_loss_logger.close()
with open(os.path.join(args.save_dir, f'duration_{args.save_id:d}.txt'),
'a') as f:
print(f'wall: {time.time() - start}')
f.write(f'wall: {time.time() - start}')
def __init__(self):
self.amazon = Amazon(api_key.amazon_key, api_key.amazon_secret)
self.price_file = "pricebg.csv"
class AmazonGet():
# Amazon のカートへ商品を入れる
def __init__(self):
self.amazon = Amazon(api_key.amazon_key, api_key.amazon_secret)
self.price_file = "pricebg.csv"
def create_cart(self, asin):
# カートを作る。
# 以降は返り値のcartidとhmacが必要
# 購入はpurchaseurlにアクセス
res = self.amazon.cartCreate(str(asin), 1)
soup = BeautifulSoup(res)
ans = []
try:
ans.append(soup.find('cartid').text.encode('sjis'))
ans.append(soup.find('hmac').text.encode('sjis'))
ans.append(soup.find('urlencodedhmac').text.encode('sjis'))
ans.append(soup.find('purchaseurl').text)
except UnicodeEncodeError:
return []
except AttributeError:
return []
return ans
def cart_add(self, asin, cartid, hmac):
# 指定された商品を一個カートに入れる
res = self.amazon.cartAdd(asin, 1, cartid, hmac)
def cart_get(self, cartid, hmac):
# カートの情報を手に入れる
# デバッグ用
res = self.amazon.cartGet(cartid, hmac)
soup = BeautifulSoup(res)
print soup.prettify()
ans = []
try:
ans.append(soup.find('subtotal').text.encode('sjis'))
except UnicodeEncodeError:
return []
except AttributeError:
return []
return ans
def solv_nap(self, nap_size, item_list):
"""
ナップサック問題を解く
価値は全て１、商品の大きさを実際の価格にしている
これにより、指定された値段内で商品を詰められる
"""
gold = [-1] * (nap_size + 1)
gold[0] = 0
for item in item_list:
for i in range(nap_size + 1):
if gold[i] != -1:
if i + item.size < nap_size:
if gold[i + item.size] < gold[i] + item.price:
gold[i + item.size] = gold[i] + item.price
# 最も最大金額に近いものを取り出す
for i in range(nap_size, 0, -1):
if gold[i] != -1:
x = i
break
# 何を買ったかを調べる
ans = []
item_list.reverse()
for item in item_list:
if 0 <= x - item.size:
if gold[x] != gold[x - item.size]:
ans.append((str(item.size), str(item.asin)))
x -= item.size
return ans
def get_amazon_url(self, money):
# 指定された金額でぴったりになるように商品を買う
# 買う為のurlを返す
reader = csv.reader(open(self.price_file, 'r'))
# 必要な情報は値段とasinなのでそれを取り出す
prices = set([])
item_list = []
for row in reader:
if int(row[1]) not in prices:
item_list.append(Item(1, int(row[1]), 0, row[0], row[2]))
prices.add(int(row[1]))
items = self.solv_nap(money, item_list)
if items != None:
id = self.create_cart(items[0][1])
for item in items[1:]:
self.cart_add(item[1], id[0], id[1])
return id[3]
def __init__(self):
self.amazon = Amazon(api_key.amazon_key, api_key.amazon_secret)
self.price_file = "pricebg.csv"
def get_amazon_data(self, query='Mobile', total_reviews=5):
from amazon import Amazon
obj = Amazon()
review, rating = obj.search(query=query, total_reviews=total_reviews)
return [review, rating]
import binascii
import os
import struct
import sys
# reset working directory to scripts directory
dir_path = os.path.dirname(os.path.realpath(__file__))
os.chdir(dir_path)
sys.stderr.write("[license_req.py] Using working directory: %s\n" % os.getcwd())
from amazon import Amazon
if __name__ == "__main__":
az = Amazon()
az.logIn()
if len(sys.argv) != 3:
raise Exception("Usage: python %s resId challenge" % sys.argv[0])
resId = sys.argv[1]
challenge = sys.argv[2]
sys.stderr.write("[license_req.py] challenge (%d): %s\n" % (len(challenge), challenge) )
serverResponse = az.getWidevine2License(resId, challenge)
assert ("widevine2License" in serverResponse)
assert ("license" in serverResponse["widevine2License"])
assert ("metadata" in serverResponse["widevine2License"])
assert ("keyMetadata" in serverResponse["widevine2License"]["metadata"])
license = serverResponse["widevine2License"]["license"]
from flipkart import Flipkart
from amazon import Amazon
from olx import Olx
product_name = input("Enter product name:- ")
# Fetch data from Flipkart and store them into flipkart.csv
f = Flipkart(product_name)
f.get_details()
f.store_data()
# Fetch data from Amazon and store them into amazon.csv
a = Amazon(product_name)
a.get_details()
a.store_data()
# Fetch data from Olx and store them into olx.csv
o = Olx(product_name)
o.get_details()
o.store_data()
# read data from CSVs files and show them on bar-plot
def get_data():
with open('amazon.csv', 'r') as d:
def main(args):
ts = time.strftime('%Y-%b-%d-%H:%M:%S', time.gmtime())
splits = ['train', 'valid'] + (['test'] if args.test else [])
datasets = OrderedDict()
for split in splits:
# datasets[split] = BGoogle(
# data_dir=args.data_dir,
# split=split,
# create_data=args.create_data,
# batch_size=args.batch_size ,
# max_sequence_length=args.max_sequence_length,
# min_occ=args.min_occ
# )
datasets[split] = Amazon(data_dir=args.data_dir,
split=split,
create_data=args.create_data,
batch_size=args.batch_size,
max_sequence_length=args.max_sequence_length,
min_occ=args.min_occ)
model = SentenceVAE(vocab_size=datasets['train'].vocab_size,
sos_idx=datasets['train'].sos_idx,
eos_idx=datasets['train'].eos_idx,
pad_idx=datasets['train'].pad_idx,
unk_idx=datasets['train'].unk_idx,
max_sequence_length=args.max_sequence_length,
embedding_size=args.embedding_size,
rnn_type=args.rnn_type,
hidden_size=args.hidden_size,
word_dropout=args.word_dropout,
embedding_dropout=args.embedding_dropout,
latent_size=args.latent_size,
num_layers=args.num_layers,
bidirectional=args.bidirectional)
if torch.cuda.is_available():
model = model.cuda()
print(model)
tokenizer = TweetTokenizer(preserve_case=False)
vocab_file = "amazon.vocab.json"
with open(os.path.join(args.data_dir, vocab_file), 'r') as file:
vocab = json.load(file)
w2i, i2w = vocab['w2i'], vocab['i2w']
if args.tensorboard_logging:
writer = SummaryWriter(
os.path.join(args.logdir, expierment_name(args, ts)))
writer.add_text("model", str(model))
writer.add_text("args", str(args))
writer.add_text("ts", ts)
# save_model_path = os.path.join(args.save_model_path, ts)
save_model_path = args.save_model_path
if not os.path.exists(save_model_path):
os.makedirs(save_model_path)
def kl_anneal_function(anneal_function, step, k, x0):
if anneal_function == 'logistic':
return float(1 / (1 + np.exp(-k * (step - x0))))
elif anneal_function == 'linear':
return min(1, step / x0)
NLL = torch.nn.NLLLoss(size_average=False,
ignore_index=datasets['train'].pad_idx)
def loss_fn(logp, target, length, mean, logv, anneal_function, step, k,
x0):
# cut-off unnecessary padding from target, and flatten
target = target[:, :torch.max(length).data].contiguous().view(-1)
logp = logp.view(-1, logp.size(2))
# Negative Log Likelihood
NLL_loss = NLL(logp, target)
# KL Divergence
KL_loss = -0.5 * torch.sum(1 + logv - mean.pow(2) - logv.exp())
KL_weight = kl_anneal_function(anneal_function, step, k, x0)
return NLL_loss, KL_loss, KL_weight
optimizer = torch.optim.Adam(model.parameters(), lr=args.learning_rate)
tensor = torch.cuda.FloatTensor if torch.cuda.is_available(
) else torch.Tensor
step = 0
save_mode = True
last_ELBO = 1e32
for epoch in range(args.epochs):
print("+" * 20)
# f_test_example(model, tokenizer, w2i, i2w)
for split in splits:
# data_loader = DataLoader(
# dataset=datasets[split],
# batch_size=args.batch_size,
# shuffle=split=='train',
# num_workers=cpu_count(),
# pin_memory=torch.cuda.is_available()
# )
batch_size = args.batch_size
tracker = defaultdict(tensor)
# Enable/Disable Dropout
if split == 'train':
model.train()
else:
model.eval()
# for iteration, batch in enumerate(data_loader):
iteration = 0
iteration_total = datasets[split].batch_num
print("batch_num", iteration_total)
for input_batch_tensor, target_batch_tensor, length_batch_tensor in datasets[
split]:
if torch.is_tensor(input_batch_tensor):
input_batch_tensor = to_var(input_batch_tensor)
if torch.is_tensor(target_batch_tensor):
target_batch_tensor = to_var(target_batch_tensor)
if torch.is_tensor(length_batch_tensor):
length_batch_tensor = to_var(length_batch_tensor)
# batch_size = batch['input'].size(0)
# for k, v in batch.items():
# if torch.is_tensor(v):
# batch[k] = to_var(v)
# Forward pass
# logp, mean, logv, z = model(batch['input'], batch['length'])
logp, mean, logv, z = model(input_batch_tensor,
length_batch_tensor)
# loss calculation
NLL_loss, KL_loss, KL_weight = loss_fn(
logp, target_batch_tensor, length_batch_tensor, mean, logv,
args.anneal_function, step, args.k, args.x0)
loss = (NLL_loss + KL_weight * KL_loss) / batch_size
# backward + optimization
if split == 'train':
optimizer.zero_grad()
loss.backward()
optimizer.step()
step += 1
iteration += 1
# bookkeepeing
# print("elbo", tracker['ELBO'])
# print("loss", loss)
if iteration == 0:
tracker['ELBO'] = loss.data
tracker['ELBO'] = tracker['ELBO'].view(1)
else:
tracker['ELBO'] = torch.cat(
(tracker['ELBO'], loss.view(1)))
if args.tensorboard_logging:
# print(loss.data)
writer.add_scalar("%s/ELBO" % split.upper(),
loss.data.item(),
epoch * iteration_total + iteration)
writer.add_scalar("%s/NLL Loss" % split.upper(),
NLL_loss.data.item() / batch_size,
epoch * iteration_total + iteration)
writer.add_scalar("%s/KL Loss" % split.upper(),
KL_loss.data.item() / batch_size,
epoch * iteration_total + iteration)
writer.add_scalar("%s/KL Weight" % split.upper(),
KL_weight,
epoch * iteration_total + iteration)
if iteration % args.print_every == 0 or iteration + 1 == iteration_total:
print(
"%s Batch %04d/%i, Loss %9.4f, NLL-Loss %9.4f, KL-Loss %9.4f, KL-Weight %6.3f"
% (split.upper(), iteration, iteration_total - 1,
loss.data.item(), NLL_loss.data.item() / batch_size,
KL_loss.data.item() / batch_size, KL_weight))
# if split == 'valid':
# if 'target_sents' not in tracker:
# tracker['target_sents'] = list()
# tracker['target_sents'] += idx2word(batch['target'].data, i2w=datasets['train'].get_i2w(), pad_idx=datasets['train'].pad_idx)
# # print("z", tracker['z'], z)
# tracker['z'] = torch.cat((tracker['z'], z.data), dim=0)
# break
print("%s Epoch %02d/%i, Mean ELBO %9.4f" %
(split.upper(), epoch, args.epochs,
torch.mean(tracker['ELBO'])))
cur_ELBO = torch.mean(tracker['ELBO'])
if args.tensorboard_logging:
writer.add_scalar("%s-Epoch/ELBO" % split.upper(), cur_ELBO,
epoch)
if split == "valid":
if cur_ELBO < last_ELBO:
save_mode = True
else:
save_mode = False
last_ELBO = cur_ELBO
# save a dump of all sentences and the encoded latent space
# if split == 'valid':
# dump = {'target_sents':tracker['target_sents'], 'z':tracker['z'].tolist()}
# if not os.path.exists(os.path.join('dumps', ts)):
# os.makedirs('dumps/'+ts)
# with open(os.path.join('dumps/'+ts+'/valid_E%i.json'%epoch), 'w') as dump_file:
# json.dump(dump,dump_file)
# save checkpoint
if split == 'train':
# checkpoint_path = os.path.join(save_model_path, "E%i.pytorch"%(epoch))
checkpoint_path = os.path.join(save_model_path, "best.pytorch")
if save_mode == True:
torch.save(model.state_dict(), checkpoint_path)
print("Model saved at %s" % checkpoint_path)
app.add_url_rule('/',
view_func=Main.as_view('main'),
methods=["GET"])
app.add_url_rule('/<page>/',
view_func=Main.as_view('page'),
methods=["GET"])
app.add_url_rule('/login/',
view_func=Login.as_view('login'),
methods=["GET","POST"])
app.add_url_rule('/register/',
view_func=Register.as_view('register'),
methods=["GET","POST"])
app.add_url_rule('/amazon/',
view_func=Amazon.as_view('amazon'),
methods=["GET","POST"])
@app.errorhandler(404)
def page_not_found(error):
return flask.render_template('404.html'), 404
# Close DB
@app.teardown_appcontext
def close_connection(exception):
db = getattr(g, '_database', None)
if db is not None:
db.close()
app.debug = True
class AmazonGet():
# Amazon のカートへ商品を入れる
def __init__(self):
self.amazon = Amazon(api_key.amazon_key, api_key.amazon_secret)
self.price_file = "pricebg.csv"
def create_cart(self,asin):
# カートを作る。
# 以降は返り値のcartidとhmacが必要
# 購入はpurchaseurlにアクセス
res = self.amazon.cartCreate(str(asin), 1)
soup = BeautifulSoup(res)
ans = []
try:
ans.append(soup.find('cartid').text.encode('sjis'))
ans.append(soup.find('hmac').text.encode('sjis'))
ans.append(soup.find('urlencodedhmac').text.encode('sjis'))
ans.append(soup.find('purchaseurl').text)
except UnicodeEncodeError:
return []
except AttributeError:
return []
return ans
def cart_add(self,asin, cartid, hmac):
# 指定された商品を一個カートに入れる
res = self.amazon.cartAdd(asin,1,cartid, hmac)
def cart_get(self, cartid, hmac):
# カートの情報を手に入れる
# デバッグ用
res = self.amazon.cartGet(cartid, hmac)
soup = BeautifulSoup(res)
print soup.prettify()
ans = []
try:
ans.append(soup.find('subtotal').text.encode('sjis'))
except UnicodeEncodeError:
return []
except AttributeError:
return []
return ans
def solv_nap(self,nap_size, item_list):
"""
ナップサック問題を解く
価値は全て１、商品の大きさを実際の価格にしている
これにより、指定された値段内で商品を詰められる
"""
gold = [-1] * (nap_size+1)
gold[0] = 0
for item in item_list:
for i in range(nap_size+1):
if gold[i] != -1:
if i+item.size < nap_size:
if gold[i+item.size] < gold[i] + item.price:
gold[i+item.size] = gold[i]+item.price
# 最も最大金額に近いものを取り出す
for i in range(nap_size, 0, -1):
if gold[i] != -1:
x = i
break
# 何を買ったかを調べる
ans = []
item_list.reverse()
for item in item_list:
if 0 <= x - item.size:
if gold[x] != gold[x-item.size]:
ans.append((str(item.size),str(item.asin)))
x -= item.size
return ans
def get_amazon_url(self,money):
# 指定された金額でぴったりになるように商品を買う
# 買う為のurlを返す
reader = csv.reader(open(self.price_file, 'r'))
# 必要な情報は値段とasinなのでそれを取り出す
prices = set([])
item_list = []
for row in reader:
if int(row[1]) not in prices:
item_list.append(Item(1,int(row[1]),0,row[0], row[2]))
prices.add(int(row[1]))
items = self.solv_nap(money, item_list)
if items != None:
id = self.create_cart(items[0][1])
for item in items[1:]:
self.cart_add(item[1], id[0], id[1])
return id[3]
def main(secondsToAudit, configFile, gracePeriod, historyFile, logFile,
auditPayments, auditRefunds):
global _config
global _civiDB
global _awsLink
global _stompLink
global _logFile
startTime = datetime.fromtimestamp(
int(time.time()) - int(secondsToAudit), pytz.utc)
endTime = datetime.fromtimestamp(
int(time.time()) - int(gracePeriod), pytz.utc)
print("AWS audit requested from %s to %s" %
(startTime.isoformat(), endTime.isoformat()))
# === Initialize the configuration file ===
localdir = os.path.dirname(os.path.abspath(__file__))
_config = SafeConfigParser()
fileList = ["%s/amazon-audit.cfg" % localdir]
if configFile is not None:
fileList.append(configFile)
_config.read(fileList)
# === Open up ze STOMP ===
_stompLink = DistStomp(_config.get('Stomp', 'server'),
_config.getint('Stomp', 'port'))
_stompLink.connect()
# === Connection to Amazon ===
_awsLink = Amazon(awsEndpoint=_config.get('AwsConfig', 'endpoint'),
awsAccessKey=_config.get('AwsConfig', 'accessKey'),
awsSecret=_config.get('AwsConfig', 'secretKey'))
# === Connection to MySQL ===
_civiDB = MySQL.connect(_config.get('MySQL', 'host'),
_config.get('MySQL', 'user'),
_config.get('MySQL', 'password'),
_config.get('MySQL', 'schema'))
# === Open up the history and log files ===
# If the history file exists, it will modify the start time of this script to be the end time of the
# history file.
hfile = None
historyStart = startTime
if historyFile and os.path.exists(historyFile):
hfile = open(historyFile, 'r')
if hfile.readline().strip() == AWS_HISTORY_FILE_VERSTR:
historyStart = dateutil.parser.parse(hfile.readline().strip())
historyEnd = dateutil.parser.parse(hfile.readline().strip())
startTime = historyEnd
print("History file modified search period, now %s to %s" %
(startTime.isoformat(), endTime.isoformat()))
else:
print('Not starting with a valid history file.')
if logFile:
_logFile = open(logFile, 'a')
_logFile.write("!!! Starting run for dates %s -> %s\n" %
(startTime.isoformat(), endTime.isoformat()))
# === Sanity checks ===
if endTime < startTime:
startTime = endTime
# === Main Application ===
awsTransactions = []
# --- Process all previously pending transactions from the history file. If the transaction is still in some form
# of pending, add it back to the history list.
historyCount = 0
historyList = []
historyStats = {'Success': 0, 'Pending': 0, 'Failed': 0, 'Ignored': 0}
if hfile:
print("Processing history file")
for txn in hfile:
historyCount += 1
awsTransactions.append(json.loads(txn))
hfile.close()
# --- Obtain AWS history ---
if auditPayments:
print("Obtaining AWS payment transactions for the period %s -> %s" %
(startTime.isoformat(), endTime.isoformat()))
awsTransactions += _awsLink.getAccountActivity(startTime,
endDate=endTime,
fpsOperation='Pay')
print("Obtained %d transactions" % len(awsTransactions))
if auditRefunds:
print("Obtaining AWS refund transactions for the period %s -> %s" %
(startTime.isoformat(), endTime.isoformat()))
awsTransactions += _awsLink.getAccountActivity(startTime,
endDate=endTime,
fpsOperation='Refund')
print("Obtained %d transactions" % len(awsTransactions))
# --- Main loop: checks each aws transaction against the Civi database; adding it if it doesn't exist ---
txncount = 0
for txn in awsTransactions:
txncount += 1
result = dispatchTransaction(txn, auditPayments, auditRefunds)
historyStats[result] += 1
if result == 'Pending':
historyList.append(txn)
print("\n--- Finished processing of messages. ---\n")
# --- Write the history file ---
if historyFile:
print("Rewriting history file with %d transactions" % len(historyList))
hfile = open(historyFile, 'w')
hfile.write("%s\n%s\n%s\n" %
(AWS_HISTORY_FILE_VERSTR, historyStart.isoformat(),
endTime.isoformat()))
for txn in historyList:
hfile.write("%s\n" % json.dumps(txn))
print("Flushing history file in preparation for main loop")
hfile.flush()
# --- Final statistics
print("Processed %d AWS messages" % txncount)
print(" ... of which %d messages were from history" % historyCount)
print("This resulted in the following:")
for entry in historyStats.items():
print(" %s Messages: %d" % entry)
# === Final Application Cleanup ===
print("\nCleaning up.")
_civiDB.close()
_stompLink.disconnect()
if hfile:
hfile.close()
if _logFile:
_logFile.close()
time.sleep(1) # Let the STOMP library catch up
def main(args):
data_name = args.data_name
with open(args.data_dir+data_name+'.vocab.json', 'r') as file:
vocab = json.load(file)
w2i, i2w = vocab['w2i'], vocab['i2w']
model = SentenceVAE(
vocab_size=len(w2i),
sos_idx=w2i['<sos>'],
eos_idx=w2i['<eos>'],
pad_idx=w2i['<pad>'],
unk_idx=w2i['<unk>'],
max_sequence_length=args.max_sequence_length,
embedding_size=args.embedding_size,
rnn_type=args.rnn_type,
hidden_size=args.hidden_size,
word_dropout=args.word_dropout,
embedding_dropout=args.embedding_dropout,
latent_size=args.latent_size,
num_layers=args.num_layers,
bidirectional=args.bidirectional
)
if not os.path.exists(args.load_checkpoint):
raise FileNotFoundError(args.load_checkpoint)
model.load_state_dict(torch.load(args.load_checkpoint))
print("Model loaded from %s"%(args.load_checkpoint))
if torch.cuda.is_available():
model = model.cuda()
model.eval()
samples, z = model.inference(n=args.num_samples)
print('----------SAMPLES----------')
print(*idx2word(samples, i2w=i2w, pad_idx=w2i['<pad>']), sep='\n')
z1 = torch.randn([args.latent_size]).numpy()
z2 = torch.randn([args.latent_size]).numpy()
z = to_var(torch.from_numpy(interpolate(start=z1, end=z2, steps=8)).float())
samples, _ = model.inference(z=z)
print('-------INTERPOLATION-------')
print(*idx2word(samples, i2w=i2w, pad_idx=w2i['<pad>']), sep='\n')
print('-------Encode ... Decode-------')
datasets = Amazon(
data_dir=args.data_dir,
split="valid",
create_data=False,
batch_size=10,
max_sequence_length=args.max_sequence_length,
min_occ=3
)
iteration = 0
for input_batch_tensor, target_batch_tensor, length_batch_tensor in datasets:
if torch.is_tensor(input_batch_tensor):
input_batch_tensor = to_var(input_batch_tensor)
if torch.is_tensor(target_batch_tensor):
target_batch_tensor = to_var(target_batch_tensor)
if torch.is_tensor(length_batch_tensor):
length_batch_tensor = to_var(length_batch_tensor)
print("*"*10)
print("->"*10, *idx2word(input_batch_tensor, i2w=i2w, pad_idx=w2i['<pad>']), sep='\n')
logp, mean, logv, z = model(input_batch_tensor,length_batch_tensor)
samples, z = model.inference(z=z)
print("<-"*10, *idx2word(samples, i2w=i2w, pad_idx=w2i['<pad>']), sep='\n')
# print("+"*10)
if iteration == 0:
break
iteration += 1
def main():
# === Extract options ===
parser = OptionParser(usage="usage: %prog [options] <# of seconds to audit>")
parser.add_option("-c", "--config", dest='configFile', default=None, help='Path to configuration file')
parser.add_option("-g", "--gracePeriod", dest='gracePeriod', default=0, help='Number of seconds from now backwards to ignore')
parser.add_option("-i", "--historyFile", dest='historyFile', default=None, help='Stores any pending transactions and the last run time')
parser.add_option('-l', "--logFile", dest='logFile', default=None, help='Saves a log of all Amazon transactions')
(options, args) = parser.parse_args()
if len(args) != 1:
parser.print_usage()
exit()
startTime = datetime.fromtimestamp(int(time.time()) - int(args[0]), pytz.utc)
endTime = datetime.fromtimestamp(int(time.time()) - int(options.gracePeriod), pytz.utc)
print("AWS refund audit requested from %s to %s" % (startTime.isoformat(), endTime.isoformat()))
# === Get the configuration options ===
config = SafeConfigParser()
fileList = ['./amazon-config.cfg']
if options.configFile is not None:
fileList.append(options.configFile)
config.read(fileList)
# === Open up ze STOMP ===
sc = DistStomp(config.get('Stomp', 'server'), config.getint('Stomp', 'port'))
sc.connect()
# === Connection to Amazon ===
aws = Amazon(
awsEndpoint = config.get('AwsConfig', 'endpoint'),
awsAccessKey = config.get('AwsConfig', 'accessKey'),
awsSecret = config.get('AwsConfig', 'secretKey')
)
# === Connection to MySQL ===
dbcon = MySQL.connect(
config.get('MySQL', 'host'),
config.get('MySQL', 'user'),
config.get('MySQL', 'password'),
config.get('MySQL', 'schema')
)
# === Open up the history and log files ===
# If the history file exists, it will modify the start time of this script to be the end time of the
# history file.
hfile = None
historyStart = startTime
historyEnd = endTime
if options.historyFile and os.path.exists(options.historyFile):
hfile = open(options.historyFile, 'r')
if hfile.readline().strip() == AWS_HISTORY_FILE_VERSTR:
historyStart = dateutil.parser.parse(hfile.readline().strip())
historyEnd = dateutil.parser.parse(hfile.readline().strip())
startTime = historyEnd
print("History file modified search period, now %s to %s" % (startTime.isoformat(), endTime.isoformat()))
else:
print('Not starting with a valid history file.')
sfile = None
if options.logFile:
sfile = open(options.logFile, 'a')
sfile.write("!!! Starting run for dates %s -> %s\n" % (startTime.isoformat(), endTime.isoformat()))
# === Sanity checks ===
if endTime < startTime:
startTime = endTime
# === Main Application ===
# --- Process all previously pending transactions from the history file. If the transaction is still in some form
# of pending, add it back to the history list.
historyCount = 0
historyList = []
historyStats = {
'Success': 0,
'Pending': 0,
'Failed': 0,
'Ignored': 0
}
if hfile:
print("Processing history file")
for txn in hfile:
historyCount += 1
txn = json.loads(txn)
result = processTransaction(txn, dbcon, aws, sc, sfile, config)
historyStats[result] += 1
if result == 'Pending':
historyList.append(txn)
hfile.close()
# --- Obtain AWS history ---
print("Obtaining AWS transactions for the period %s -> %s" % (startTime.isoformat(), endTime.isoformat()))
awsTransactions = aws.getAccountActivity(startTime, endDate=endTime, fpsOperation='Pay')
print("Obtained %d transactions" % len(awsTransactions))
# --- Main loop: checks each aws transaction against the Civi database; adding it if it doesn't exist ---
txncount = 0
for txn in awsTransactions:
txncount += 1
result = processTransaction(txn, dbcon, aws, sc, sfile, config)
historyStats[result] += 1
if result == 'Pending':
historyList.append(txn)
print("\n--- Finished processing of messages. ---\n")
# --- Prepare the history file for write ---
if options.historyFile:
print("Rewriting history file with %d transactions" % len(historyList))
hfile = open(options.historyFile, 'w')
hfile.write("%s\n%s\n%s\n" % (AWS_HISTORY_FILE_VERSTR, historyStart.isoformat(), endTime.isoformat()))
for txn in historyList:
hfile.write("%s\n" % json.dumps(txn))
print("Flushing history file in preparation for main loop")
hfile.flush()
# --- Final statistics ---
print("%d new AWS messages" % txncount)
print(" Additionally %d messages were processed from history" % historyCount)
print("This resulted in the following:")
for entry in historyStats.items():
print(" %s Messages: %d" % entry)
# === Final Application Cleanup ===
print("\nCleaning up.")
dbcon.close()
sc.disconnect()
if hfile:
hfile.close()
if sfile:
sfile.close()
time.sleep(1) # Let the STOMP library catch up