def fnctot_mean(input_file_list, dist_num):
split_file_list = utils.chunkIt(input_file_list, dist_num)
queue_list = []
procs = []
for i, file_list in enumerate(split_file_list):
queue_list.append(
Queue()) # define queues for saving the outputs of functions
procs.append(
Process(target=get_stft_mean,
args=(file_list, queue_list[i]))) # define process
for p in procs: # process start
p.start()
M_list = []
for i in range(dist_num): # save results from queues and close queues
M_list.append(queue_list[i].get())
queue_list[i].close()
for p in procs: # close process
p.terminate()
return M_list
def run(self):
self.m.createBucket(conf.new_bucket_name)
packet_names = getJsonFile(packetListPath)
myPrint("Total " + str(len(packet_names)) + " packets found", 12)
if conf.records is not None:
packet_names = packet_names[0:conf.records]
packet_names_chunks = chunkIt(packet_names, conf.threads)
i = 0
for packet_names_chunk in packet_names_chunks:
myPrint("Chunk " + str(i + 1) +
": total packet to be migrated are " +
str(len(packet_names_chunk)))
pool = Pool(conf.threads)
pool.map(runner, packet_names_chunks)
def lpsd_dist_p(self, data, dist_num, is_patch=True):
data = data.tolist()
chunk_list = utils.chunkIt(data, dist_num)
data_list = []
for item in chunk_list:
if len(item) > 1:
data_list.append(item)
while True:
queue_list = []
procs = []
for queue_val in queue_list:
print(queue_val.empty())
for i, data in enumerate(data_list):
queue_list.append(Queue(
)) # define queues for saving the outputs of functions
procs.append(
Process(target=self.get_lpsd_p,
args=(data, queue_list[i]))) # define process
for queue_val in queue_list:
print(queue_val.empty())
for p in procs: # process start
p.start()
for queue_val in queue_list:
while queue_val.empty():
pass
for queue_val in queue_list:
print(queue_val.empty())
M_list = []
for i in range(
dist_num): # save results from queues and close queues
# while not queue_list[i].empty():
get_time = time.time()
M_list.append(queue_list[i].get(timeout=3))
get_time = time.time() - get_time
queue_list[i].close()
queue_list[i].join_thread()
for p in procs: # close process
p.terminate()
if get_time < 3:
break
else:
print('Some error occurred, restarting the lpsd extraction...')
result = np.concatenate(M_list, axis=0)
# result = np.asarray(M_list)
# print(result.shape)
# result = np.reshape(result, (-1, result.shape[2], result.shape[3]))
lpsd = np.expand_dims(
result[:, :, 0], axis=2
) # expand_dims for normalization (shape matching for broadcast)
# if not self._is_training:
self.phase.append(result[:, :, 1])
pad = np.expand_dims(np.zeros(
(int(config.time_width / 2), lpsd.shape[1])),
axis=2) # pad for extracting the patches
if is_patch:
mean, std = self.norm_process(self._norm_dir + '/norm_noisy.mat')
lpsd = self._normalize(mean, std, lpsd)
else:
mean, std = self.norm_process(self._norm_dir + '/norm_noisy.mat')
lpsd = self._normalize(mean, std, lpsd)
# print(result.shape)
if is_patch:
lpsd = np.squeeze(np.concatenate((pad, lpsd, pad),
axis=0)) # padding for patching
# print(result.shape)
lpsd = image.extract_patches_2d(lpsd,
(config.time_width, lpsd.shape[1]))
lpsd = np.expand_dims(lpsd, axis=3)
return lpsd
def lpsd_dist(self, data, dist_num, is_patch=True):
data = data.tolist()
data_list = utils.chunkIt(data, dist_num)
queue_list = []
procs = []
for i, data in enumerate(data_list):
queue_list.append(
Queue()) # define queues for saving the outputs of functions
procs.append(
Process(target=self.get_lpsd,
args=(data, queue_list[i]))) # define process
for p in procs: # process start
p.start()
M_list = []
for i in range(dist_num): # save results from queues and close queues
M_list.append(queue_list[i].get())
queue_list[i].close()
for p in procs: # close process
p.join()
result = np.concatenate(M_list, axis=0)
# result = np.asarray(M_list)
# print(result.shape)
# result = np.reshape(result, (-1, result.shape[2], result.shape[3]))
lpsd = np.expand_dims(
result[:, :, 0], axis=2
) # expand_dims for normalization (shape matching for broadcast)
if not self._is_training:
self.phase.append(result[:, :, 1])
pad = np.expand_dims(np.zeros(
(int(config.time_width / 2), lpsd.shape[1])),
axis=2) # pad for extracting the patches
if is_patch:
mean, std = self.norm_process(self._norm_dir + '/norm_noisy.mat')
lpsd = self._normalize(mean, std, lpsd)
else:
mean, std = self.norm_process(self._norm_dir + '/norm_noisy.mat')
lpsd = self._normalize(mean, std, lpsd)
# print(result.shape)
if is_patch:
lpsd = np.squeeze(np.concatenate((pad, lpsd, pad),
axis=0)) # padding for patching
# print(result.shape)
lpsd = image.extract_patches_2d(lpsd,
(config.time_width, lpsd.shape[1]))
lpsd = np.expand_dims(lpsd, axis=3)
return lpsd
