def update_job_status(self, job_id):
"""Updates a spark job's status based on information from getJobInfo()."""
v = self.jobs[job_id]
jobinfo = self.status.getJobInfo(v["spark_id"])
done_string = 'false' if jobinfo.status == 'RUNNING' else 'true'
total_total = v["chunk_size"] * v["chunks"]
if v["finished"] == v["started"] and done_string == "true":
v["chunks_done"] += 1
if v["chunks_done"] == v["chunks"]:
v["finished"] = datetime.now()
total_done = v["chunk_size"] * v["chunks_done"]
else:
(nTasks, nActive, nComplete) = (0, 0, 0)
for sid in jobinfo.stageIds:
stageinfo = self.status.getStageInfo(sid)
if stageinfo:
nTasks += stageinfo.numTasks
nActive += stageinfo.numActiveTasks
nComplete += stageinfo.numCompletedTasks
if v["chunks"] > 0 and v["chunk_size"] == 0:
v["chunk_size"] = nTasks
total_done = v["chunk_size"] * v["chunks_done"] + nComplete
total_done = min(total_done, total_total)
my_print("Setting job totals: %d %d %d %d %d" %
(v["chunk_size"], v["chunks"], v["chunks_done"], total_total,
total_done))
self.db.query('''
UPDATE jobs SET tasks_done=%d, tasks_total=%d, status='%s', done=%s, finish='%s'
WHERE id=%d
''' % (total_done, total_total, jobinfo.status, done_string,
str(self.jobs[job_id]["finished"]), job_id))
def get_webpage_content(self, url):
# return str html
try:
req = urllib.request.Request(url)
response = urllib.request.urlopen(req)
webpage_str = response.read()
return str(webpage_str)
except Exception as e:
my_print("error reading:" + url)
return None
def update_all_jobs_callback(self):
"""For each job, checks whether its status has changed."""
try:
my_print("updating spark jobs status...")
for job_id, v in self.jobs.iteritems():
if v["finished"] == v["started"]:
self.update_job_status(job_id)
finally:
tornado.ioloop.IOLoop.instance().add_timeout(
timedelta(seconds=5), self.update_all_jobs_callback)
def get_acc(param):
if len(data.training_input) < config.training_segment:
print_error("training segment > data number")
return False
d = 0.0
# Run all seq between [training_segment, training_input_len)
for i in range(config.training_segment, data.training_input_len + 1, RNN.config.batch_num):
c = data.training_input_sequential_selection(RNN.config.batch_num, i)
d += RNN.train.cost(c[0], c[1])
# print(RNN.train.dimshuffle(c[0]))
my_print("Cost", d)
def train(self, X, Y,
ordered_topics,
balance='p',
max_per_class=100):
topic = ''
for k in xrange(self.K):
topic = ordered_topics[k]
Y_k = Y[:,k:k+1]
self.svms[k].train(X, Y_k,
balance=balance,
max_per_class=max_per_class,
topic=topic)
util.my_print('%d . . .' % (k+1), same_line=True)
def get_lastest_page_info(self, latest_webpage_str):
my_print("get page info")
date = re.findall(r"Date:\s+(.*),\s+2020<br/>", latest_webpage_str)
positive_cases = re.findall(r"([,\d]+)[\\\d\w\s]+Positive cases", latest_webpage_str)
death = re.findall(r"([,\d]+)[\\\d\w\s]+Deaths", latest_webpage_str)
pending = re.findall(r" least ([\+,\d]+) results have been received and another ([\+,\d]+) are pending", latest_webpage_str)
return {
'date': date[0] if len(date) > 0 else None,
'positive_cases': positive_cases[0] if len(positive_cases) > 0 else None,
'death': death[0] if len(death) > 0 else None,
'results received': pending[0][0] if len(pending) else None,
'results pending': pending[0][1] if len(pending) else None
}
def main():
"""Creates tornado application, handles keyboard interrupts (to release the http socket)."""
global args, config
parser = argparse.ArgumentParser(description='IMS webserver.')
parser.add_argument('--config',
dest='config',
type=str,
help='config file name')
parser.add_argument('--port',
dest='port',
type=int,
help='port on which to access the web server')
parser.add_argument('--profile',
dest='time_profiling_enabled',
action='store_true')
parser.add_argument('--use-deprecated',
dest='use_deprecated',
action='store_true')
parser.add_argument('--debug', dest='debug', action='store_true')
parser.set_defaults(spark=False,
config='config.json',
port=8080,
time_profiling_enabled=False,
use_deprecated=False)
args = parser.parse_args()
# handlers.args = args
with open(args.config) as f:
config = json.load(f)
port = args.port
torn_app = Application(args.debug)
http_server = tornado.httpserver.HTTPServer(torn_app)
http_server.listen(port)
my_print("Starting server, listening to port %d..." % port)
try:
# set periodic updates
if args.spark:
tornado.ioloop.IOLoop.instance().add_timeout(
timedelta(seconds=5), torn_app.update_all_jobs_callback)
# start loop
tornado.ioloop.IOLoop.instance().start()
except KeyboardInterrupt:
my_print('^C received, shutting down server')
if args.spark:
torn_app.sc.stop()
http_server.stop()
def __init__(self, args, qry_file, printDet=False):
self.queries, self.thres_10, self.thres_100, self.thres_1000, self.qids, self.qterms, _ = read_queries_and_thres(
qry_file, 0, printDet)
self.tensor_queries = create_tensors_from_np(self.queries)
self.qlens = []
for qt in self.qterms:
self.qlens.append(len(qt))
if args.k == 10:
self.tensor_thres = create_thresholds(self.thres_10)
if args.k == 100:
self.tensor_thres = create_thresholds(self.thres_100)
if args.k == 1000:
self.tensor_thres = create_thresholds(self.thres_1000)
my_print("dataset statistics:", qry_file)
my_print("\tqueries =", len(self.queries))
def main():
scrapper = Scrapper()
# find all occurences of /Programs/OPA/Pages/NR.*.aspx
all_urls = re.findall(r"(/Programs/OPA/Pages/NR20-\d+.aspx)",
scrapper.get_webpage_content(main_page_url))
all_urls = sorted(all_urls, key=cmp_to_key(scrapper.get_latest_date))
# get latest page content
latest = all_urls[-3]
latest_webpage_str = scrapper.get_webpage_content(website_url + latest)
dedup_url = {}
for url in all_urls[-10:]:
if url not in dedup_url:
dedup_url[url] = True
my_print(url)
content = scrapper.get_webpage_content(website_url + url)
if content:
# print(content)
print(scrapper.get_lastest_page_info(content))
def train(model, epoch, data, quant):
dataloader = DataLoader(data, batch_size=args.batch_size, shuffle=True)
quantiles = torch.tensor([[quant]]).view(1, -1).float().to(args.device)
model.train()
optim = torch.optim.Adam(model.parameters(), lr=lr)
my_print("epoch {} start training {} instances with lr {}".format(
epoch, len(dataloader), lr))
with tqdm(total=len(dataloader), unit='batches', desc='train') as pbar:
losses = []
for batch_num, batch in enumerate(dataloader):
optim.zero_grad()
queries, thres = batch
scores = model(queries.to(args.device))
loss = quantile_loss(scores, thres.to(args.device), quantiles)
losses.append(loss.item())
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), args.clip)
optim.step()
pbar.set_postfix(loss=np.mean(losses[-128:]))
pbar.update(1)
pbar.close()
help='quantiles',
type=float,
required=True)
parser.add_argument('--device',
default="cpu",
type=str,
required=False,
help='compute device')
parser.add_argument('--debug',
default=False,
dest='debug',
action='store_true')
args = parser.parse_args()
init_log(args)
torch.set_num_threads(hyperparams.default_threads)
my_print("Parameters:")
for k, v in sorted(vars(args).items()):
my_print("\t{0}: {1}".format(k, v))
# Set the random seed manually for reproducibility.
torch.manual_seed(hyperparams.random_seed)
###############################################################################
# Load data
###############################################################################
if torch.cuda.is_available():
if args.device != "cpu":
torch.cuda.set_device(int(args.device.split(":")[1]))
args.device = torch.device(args.device)
else:
args.device = torch.device('cuda')
def print_status(param):
if len(param) == 0 or param[0] == "testbench":
print("========== TestBench Status ==========")
my_print("Train input file", tc.train_input_file)
my_print("Train answer file", tc.train_answer_file)
my_print("Test input file", tc.test_input_file)
my_print("Train input file", tc.test_output_file)
if len(param) == 0 or param[0] == "rnn":
print("========== RNN Status ================")
my_print("Input dimension", rc.input_dim)
my_print("Output dimension", rc.output_dim)
my_print("Hidden layer list", rc.hidden_layer_dim_list)
my_print("Batch size", rc.batch_num)
my_print("Learning rate", rc.learning_rate)
print("======================================")
