def validate_insert(_table_name):
milvus = Milvus()
milvus.connect(host="localhost", port="19530")
status, count = milvus.get_table_row_count(_table_name)
assert count == vector_num * process_num, f"Error: validate insert not pass: "******"{vector_num * process_num} expected but {count} instead!"
milvus.disconnect()
def main():
milvus = Milvus()
milvus.connect(host=_HOST, port=_PORT)
#
# table_name = 'test_search_in_file'
# dimension = 256
# vectors = Prepare.records([[random.random()for _ in range(dimension)] for _ in range(20)])
# param = {
# 'table_name': table_name,
# 'file_ids': ['1'],
# 'query_records': vectors,
# 'top_k': 5,
# # 'query_ranges': [] # Not fully tested yet
# }
# status, result = milvus.search_vectors_in_files(**param)
# if status.OK():
# pprint(result)
#
# _, result = milvus.get_table_row_count(table_name)
# print('# Count: {}'.format(result))
table_name = 'test_search'
dimension = 256
# param = {'start_date': '2019-06-24', 'end_date': '2019-06-25'}
ranges = [['2019-06-25', '2019-06-25']]
vectors = Prepare.records([[random.random() for _ in range(dimension)]
for _ in range(1)])
# ranges = [Prepare.range(**param)]
LOGGER.info(ranges)
param = {
'table_name': table_name,
'query_records': vectors,
'top_k': 5,
'query_ranges': ranges # Not fully tested yet
}
status, result = milvus.search_vectors(**param)
if status.OK():
pprint(result)
_, result = milvus.get_table_row_count(table_name)
print('# Count: {}'.format(result))
milvus.disconnect()
def _test_add_vector_with_multiprocessing(self, args):
'''
target: test add vectors, with multi processes
method: 10 processed add vectors concurrently
expected: status ok and result length is equal to the length off added vectors
'''
table = gen_unique_str("test_add_vector_with_multiprocessing")
uri = "tcp://%s:%s" % (args["ip"], args["port"])
param = {
'table_name': table,
'dimension': dim,
'index_file_size': index_file_size
}
# create table
milvus = Milvus()
milvus.connect(uri=uri)
milvus.create_table(param)
vector = gen_single_vector(dim)
process_num = 4
loop_num = 10
processes = []
# with dependent connection
def add(milvus):
i = 0
while i < loop_num:
status, ids = milvus.add_vectors(table, vector)
i = i + 1
for i in range(process_num):
milvus = Milvus()
milvus.connect(uri=uri)
p = Process(target=add, args=(milvus, ))
processes.append(p)
p.start()
time.sleep(0.2)
for p in processes:
p.join()
time.sleep(3)
status, count = milvus.get_table_row_count(table)
assert count == process_num * loop_num
file_name = '/poc/yuncong/ann_1000m/query.npy'
data = np.load(file_name)
vectors = data[0:nq].tolist()
# print(vectors)
connect = Milvus()
# connect.connect(host="192.168.1.27")
# print(connect.show_tables())
# print(connect.get_table_row_count(table))
# sys.exit()
connect.connect(host="127.0.0.1")
connect.delete_table(table)
# sys.exit()
# time.sleep(2)
print(connect.get_table_row_count(table))
param = {
'table_name': table,
'dimension': dim,
'metric_type': MetricType.L2,
'index_file_size': 10
}
status = connect.create_table(param)
print(status)
print(connect.get_table_row_count(table))
# add vectors
for i in range(10):
status, ids = connect.add_vectors(table, vectors)
print(status)
print(ids[0])
# print(ids[0])
from milvus import Milvus, Prepare, IndexType, Status
import random
milvus = Milvus()
# Connect Milvus server, please change HOST and PORT to correct one
milvus.connect(host='localhost', port='33001')
# Table name is defined
table_name = 'table_' + str(random.randint(0, 100))
# Create table: table name, vector dimension and index type
milvus.create_table(
Prepare.table_schema(table_name, dimension=256,
index_type=IndexType.IDMAP))
# Add 20 256-dim-vectors into table
vectors = Prepare.records([[random.random() for _ in range(256)]
for _ in range(20)])
milvus.add_vectors(table_name=table_name, records=vectors)
# Get table row count
_, result = milvus.get_table_row_count(table_name=table_name)
print('Table {}, row counts: {}'.format(table_name, result))
class MilvusClient(object):
def __init__(self, table_name=None, ip=None, port=None):
self._milvus = Milvus()
self._table_name = table_name
try:
if not ip:
self._milvus.connect(
host = SERVER_HOST_DEFAULT,
port = SERVER_PORT_DEFAULT)
else:
self._milvus.connect(
host = ip,
port = port)
except Exception as e:
raise e
def __str__(self):
return 'Milvus table %s' % self._table_name
def check_status(self, status):
if not status.OK():
logger.error(status.message)
raise Exception("Status not ok")
def create_table(self, table_name, dimension, index_file_size, metric_type):
if not self._table_name:
self._table_name = table_name
if metric_type == "l2":
metric_type = MetricType.L2
elif metric_type == "ip":
metric_type = MetricType.IP
else:
logger.error("Not supported metric_type: %s" % metric_type)
create_param = {'table_name': table_name,
'dimension': dimension,
'index_file_size': index_file_size,
"metric_type": metric_type}
status = self._milvus.create_table(create_param)
self.check_status(status)
@time_wrapper
def insert(self, X, ids=None):
status, result = self._milvus.add_vectors(self._table_name, X, ids)
self.check_status(status)
return status, result
@time_wrapper
def create_index(self, index_type, nlist):
if index_type == "flat":
index_type = IndexType.FLAT
elif index_type == "ivf_flat":
index_type = IndexType.IVFLAT
elif index_type == "ivf_sq8":
index_type = IndexType.IVF_SQ8
elif index_type == "nsg":
index_type = IndexType.NSG
elif index_type == "ivf_sq8h":
index_type = IndexType.IVF_SQ8H
elif index_type == "ivf_pq":
index_type = IndexType.IVF_PQ
index_params = {
"index_type": index_type,
"nlist": nlist,
}
logger.info("Building index start, table_name: %s, index_params: %s" % (self._table_name, json.dumps(index_params)))
status = self._milvus.create_index(self._table_name, index=index_params)
self.check_status(status)
def describe_index(self):
return self._milvus.describe_index(self._table_name)
def drop_index(self):
logger.info("Drop index: %s" % self._table_name)
return self._milvus.drop_index(self._table_name)
@time_wrapper
def query(self, X, top_k, nprobe):
status, result = self._milvus.search_vectors(self._table_name, top_k, nprobe, X)
self.check_status(status)
return status, result
def count(self):
return self._milvus.get_table_row_count(self._table_name)[1]
def delete(self, timeout=60):
logger.info("Start delete table: %s" % self._table_name)
self._milvus.delete_table(self._table_name)
i = 0
while i < timeout:
if self.count():
time.sleep(1)
i = i + 1
continue
else:
break
if i < timeout:
logger.error("Delete table timeout")
def describe(self):
return self._milvus.describe_table(self._table_name)
def show_tables(self):
return self._milvus.show_tables()
def exists_table(self):
status, res = self._milvus.has_table(self._table_name)
self.check_status(status)
return res
@time_wrapper
def preload_table(self):
return self._milvus.preload_table(self._table_name, timeout=3000)
def main():
milvus = Milvus()
# Connect to Milvus server
# You may need to change _HOST and _PORT accordingly
param = {'host': _HOST, 'port': _PORT}
status = milvus.connect(**param)
# Create table demo_table if it dosen't exist.
table_name = 'demo_table_01'
if not milvus.has_table(table_name):
param = {
'table_name': table_name,
'dimension': 16,
'index_file_size': 1024,
'metric_type': MetricType.L2
}
milvus.create_table(param)
# Show tables in Milvus server
_, tables = milvus.show_tables()
# Describe demo_table
_, table = milvus.describe_table(table_name)
# 10 vectors with 16 dimension
vectors = [
[
0.66, 0.01, 0.29, 0.64, 0.75, 0.94, 0.26, 0.79, 0.61, 0.11, 0.25,
0.50, 0.74, 0.37, 0.28, 0.63
],
[
0.77, 0.65, 0.57, 0.68, 0.29, 0.93, 0.17, 0.15, 0.95, 0.09, 0.78,
0.37, 0.76, 0.21, 0.42, 0.15
],
[
0.61, 0.38, 0.32, 0.39, 0.54, 0.93, 0.09, 0.81, 0.52, 0.30, 0.20,
0.59, 0.15, 0.27, 0.04, 0.37
],
[
0.33, 0.03, 0.87, 0.47, 0.79, 0.61, 0.46, 0.77, 0.62, 0.70, 0.85,
0.01, 0.30, 0.41, 0.74, 0.98
],
[
0.19, 0.80, 0.03, 0.75, 0.22, 0.49, 0.52, 0.91, 0.40, 0.91, 0.79,
0.08, 0.27, 0.16, 0.07, 0.24
],
[
0.44, 0.36, 0.16, 0.88, 0.30, 0.79, 0.45, 0.31, 0.45, 0.99, 0.15,
0.93, 0.37, 0.25, 0.78, 0.84
],
[
0.33, 0.37, 0.59, 0.66, 0.76, 0.11, 0.19, 0.38, 0.14, 0.37, 0.97,
0.50, 0.08, 0.69, 0.16, 0.67
],
[
0.68, 0.97, 0.20, 0.13, 0.30, 0.16, 0.85, 0.21, 0.26, 0.17, 0.81,
0.96, 0.18, 0.40, 0.13, 0.74
],
[
0.11, 0.26, 0.44, 0.91, 0.89, 0.79, 0.98, 0.91, 0.09, 0.45, 0.07,
0.88, 0.71, 0.35, 0.97, 0.41
],
[
0.17, 0.54, 0.61, 0.58, 0.25, 0.63, 0.65, 0.71, 0.26, 0.80, 0.28,
0.77, 0.69, 0.02, 0.63, 0.60
],
]
# Insert vectors into demo_table
status, ids = milvus.add_vectors(table_name=table_name, records=vectors)
# Wait for 6 seconds, since Milvus server persist vector data every 5 seconds by default.
# You can set data persist interval in Milvus config file.
time.sleep(6)
# Get demo_table row count
status, result = milvus.get_table_row_count(table_name)
# Use the 3rd vector for similarity search
query_vectors = [vectors[3]]
# execute vector similarity search
param = {
'table_name': table_name,
'query_records': query_vectors,
'top_k': 1,
'nprobe': 16
}
status, results = milvus.search_vectors(**param)
if results[0][0].distance == 0.0 or results[0][0].id == ids[3]:
print('Query result is correct')
else:
print('Query result isn\'t correct')
# Delete demo_table
status = milvus.delete_table(table_name)
# Disconnect from Milvus
status = milvus.disconnect()
def main():
milvus = Milvus()
# Print client version
print('# Client version: {}'.format(milvus.client_version()))
# Connect milvus server
# Please change HOST and PORT to the correct one
param = {'host': _HOST, 'port': _PORT}
cnn_status = milvus.connect(**param)
print('# Connect Status: {}'.format(cnn_status))
# Check if connected
# is_connected = milvus.connected
print('# Is connected: {}'.format(milvus.connected))
# Print milvus server version
print('# Server version: {}'.format(milvus.server_version()))
# Describe table
table_name = 'table01'
res_status, table = milvus.describe_table(table_name)
print('# Describe table status: {}'.format(res_status))
print('# Describe table:{}'.format(table))
# Create table
# Check if `table01` exists, if not, create a table `table01`
dimension = 256
if not table:
param = {
'table_name': table_name,
'dimension': dimension,
'index_type': IndexType.IDMAP,
'store_raw_vector': False
}
res_status = milvus.create_table(Prepare.table_schema(**param))
print('# Create table status: {}'.format(res_status))
# Show tables and their description
status, tables = milvus.show_tables()
pprint(tables)
# Add vectors
# Prepare vector with 256 dimension
vectors = Prepare.records([[random.random() for _ in range(dimension)]
for _ in range(20)])
# Insert vectors into table 'table01'
status, ids = milvus.add_vectors(table_name=table_name, records=vectors)
print('# Add vector status: {}'.format(status))
pprint(ids)
# Search vectors
# When adding vectors for the first time, server will take at least 5s to
# persist vector data, so you have to wait for 6s after adding vectors for
# the first time.
print('# Waiting for 6s...')
time.sleep(6)
q_records = Prepare.records([[random.random() for _ in range(dimension)]
for _ in range(2)])
param = {
'table_name': table_name,
'query_records': q_records,
'top_k': 10,
}
status, results = milvus.search_vectors(**param)
print('# Search vectors status: {}'.format(status))
pprint(results)
# Get table row count
status, result = milvus.get_table_row_count(table_name)
print('# Status: {}'.format(status))
print('# Count: {}'.format(result))
# Disconnect
status = milvus.disconnect()
print('# Disconnect Status: {}'.format(status))
class MilvusClient(object):
def __init__(self, table_name=None, ip=None, port=None, timeout=60):
self._milvus = Milvus()
self._table_name = table_name
try:
i = 1
start_time = time.time()
if not ip:
self._milvus.connect(host=SERVER_HOST_DEFAULT,
port=SERVER_PORT_DEFAULT)
else:
# retry connect for remote server
while time.time() < start_time + timeout:
try:
self._milvus.connect(host=ip, port=port)
if self._milvus.connected() is True:
logger.debug(
"Try connect times: %d, %s" %
(i, round(time.time() - start_time, 2)))
break
except Exception as e:
logger.debug("Milvus connect failed")
i = i + 1
except Exception as e:
raise e
def __str__(self):
return 'Milvus table %s' % self._table_name
def check_status(self, status):
if not status.OK():
logger.error(status.message)
# raise Exception("Status not ok")
def create_table(self, table_name, dimension, index_file_size,
metric_type):
if not self._table_name:
self._table_name = table_name
if metric_type == "l2":
metric_type = MetricType.L2
elif metric_type == "ip":
metric_type = MetricType.IP
elif metric_type == "jaccard":
metric_type = MetricType.JACCARD
elif metric_type == "hamming":
metric_type = MetricType.HAMMING
else:
logger.error("Not supported metric_type: %s" % metric_type)
create_param = {
'table_name': table_name,
'dimension': dimension,
'index_file_size': index_file_size,
"metric_type": metric_type
}
status = self._milvus.create_table(create_param)
self.check_status(status)
@time_wrapper
def insert(self, X, ids=None):
status, result = self._milvus.add_vectors(self._table_name, X, ids)
self.check_status(status)
return status, result
@time_wrapper
def create_index(self, index_type, nlist):
index_params = {
"index_type": INDEX_MAP[index_type],
"nlist": nlist,
}
logger.info("Building index start, table_name: %s, index_params: %s" %
(self._table_name, json.dumps(index_params)))
status = self._milvus.create_index(self._table_name,
index=index_params)
self.check_status(status)
def describe_index(self):
status, result = self._milvus.describe_index(self._table_name)
index_type = None
for k, v in INDEX_MAP.items():
if result._index_type == v:
index_type = k
break
nlist = result._nlist
res = {"index_type": index_type, "nlist": nlist}
return res
def drop_index(self):
logger.info("Drop index: %s" % self._table_name)
return self._milvus.drop_index(self._table_name)
@time_wrapper
def query(self, X, top_k, nprobe):
status, result = self._milvus.search_vectors(self._table_name, top_k,
nprobe, X)
self.check_status(status)
return result
def count(self):
return self._milvus.get_table_row_count(self._table_name)[1]
def delete(self, timeout=60):
logger.info("Start delete table: %s" % self._table_name)
self._milvus.delete_table(self._table_name)
i = 0
while i < timeout:
if self.count():
time.sleep(1)
i = i + 1
continue
else:
break
if i >= timeout:
logger.error("Delete table timeout")
def describe(self):
return self._milvus.describe_table(self._table_name)
def show_tables(self):
return self._milvus.show_tables()
def exists_table(self, table_name=None):
if table_name is None:
table_name = self._table_name
status, res = self._milvus.has_table(table_name)
self.check_status(status)
return res
@time_wrapper
def preload_table(self):
return self._milvus.preload_table(self._table_name, timeout=3000)
def get_server_version(self):
status, res = self._milvus.server_version()
return res
def get_server_mode(self):
return self.cmd("mode")
def get_server_commit(self):
return self.cmd("build_commit_id")
def get_server_config(self):
return json.loads(self.cmd("get_config *"))
def get_mem_info(self):
result = json.loads(self.cmd("get_system_info"))
result_human = {
# unit: Gb
"memory_used":
round(int(result["memory_used"]) / (1024 * 1024 * 1024), 2)
}
return result_human
def cmd(self, command):
status, res = self._milvus._cmd(command)
logger.info("Server command: %s, result: %s" % (command, res))
self.check_status(status)
return res
class MilvusClient(object):
def __init__(self, table_name=None, host=None, port=None):
self._milvus = Milvus()
self._table_name = table_name
try:
if not host:
self._milvus.connect(host=SERVER_HOST_DEFAULT,
port=SERVER_PORT_DEFAULT)
else:
self._milvus.connect(host=host, port=port)
except Exception as e:
raise e
def __str__(self):
return 'Milvus table %s' % self._table_name
def check_status(self, status):
if not status.OK():
logger.error(status.message)
raise Exception("Status not ok")
def create_table(self, table_name, dimension, index_file_size,
metric_type):
if not self._table_name:
self._table_name = table_name
if metric_type == "l2":
metric_type = MetricType.L2
elif metric_type == "ip":
metric_type = MetricType.IP
else:
logger.error("Not supported metric_type: %s" % metric_type)
self._metric_type = metric_type
create_param = {
'table_name': table_name,
'dimension': dimension,
'index_file_size': index_file_size,
"metric_type": metric_type
}
status = self._milvus.create_table(create_param)
self.check_status(status)
@time_wrapper
def insert(self, X, ids):
if self._metric_type == MetricType.IP:
logger.info("Set normalize for metric_type: Inner Product")
X = sklearn.preprocessing.normalize(X, axis=1, norm='l2')
X = X.astype(numpy.float32)
status, result = self._milvus.add_vectors(self._table_name,
X.tolist(),
ids=ids)
self.check_status(status)
return status, result
@time_wrapper
def create_index(self, index_type, nlist):
if index_type == "flat":
index_type = IndexType.FLAT
elif index_type == "ivf_flat":
index_type = IndexType.IVFLAT
elif index_type == "ivf_sq8":
index_type = IndexType.IVF_SQ8
elif index_type == "ivf_sq8h":
index_type = IndexType.IVF_SQ8H
elif index_type == "nsg":
index_type = IndexType.NSG
elif index_type == "ivf_pq":
index_type = IndexType.IVF_PQ
index_params = {
"index_type": index_type,
"nlist": nlist,
}
logger.info("Building index start, table_name: %s, index_params: %s" %
(self._table_name, json.dumps(index_params)))
status = self._milvus.create_index(self._table_name,
index=index_params,
timeout=6 * 3600)
self.check_status(status)
def describe_index(self):
return self._milvus.describe_index(self._table_name)
def drop_index(self):
logger.info("Drop index: %s" % self._table_name)
return self._milvus.drop_index(self._table_name)
@time_wrapper
def query(self, X, top_k, nprobe):
if self._metric_type == MetricType.IP:
logger.info("Set normalize for metric_type: Inner Product")
X = sklearn.preprocessing.normalize(X, axis=1, norm='l2')
X = X.astype(numpy.float32)
status, results = self._milvus.search_vectors(self._table_name, top_k,
nprobe, X.tolist())
self.check_status(status)
ids = []
for result in results:
tmp_ids = []
for item in result:
tmp_ids.append(item.id)
ids.append(tmp_ids)
return ids
def count(self):
return self._milvus.get_table_row_count(self._table_name)[1]
def delete(self, table_name):
logger.info("Start delete table: %s" % table_name)
return self._milvus.delete_table(table_name)
def describe(self):
return self._milvus.describe_table(self._table_name)
def exists_table(self, table_name):
return self._milvus.has_table(table_name)
def get_server_version(self):
status, res = self._milvus.server_version()
self.check_status(status)
return res
@time_wrapper
def preload_table(self):
return self._milvus.preload_table(self._table_name)
