1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
| import argparse
import queue
import threading
from functools import partial
import grpc
import inference_pb2
import inference_pb2_grpc
import management_pb2
import management_pb2_grpc
def get_inference_stub():
channel = grpc.insecure_channel("localhost:7070")
stub = inference_pb2_grpc.InferenceAPIsServiceStub(channel)
return stub
def get_management_stub():
channel = grpc.insecure_channel("localhost:7071")
stub = management_pb2_grpc.ManagementAPIsServiceStub(channel)
return stub
def infer(stub, model_name, model_input, metadata):
with open(model_input, "rb") as f:
data = f.read()
input_data = {"data": data}
response = stub.Predictions(
inference_pb2.PredictionsRequest(model_name=model_name, input=input_data),
metadata=metadata,
)
try:
prediction = response.prediction.decode("utf-8")
print(prediction)
except grpc.RpcError as e:
print(f"gRPC error: {e.details()}")
exit(1)
def infer_stream(stub, model_name, model_input, metadata):
with open(model_input, "rb") as f:
data = f.read()
input_data = {"data": data}
responses = stub.StreamPredictions(
inference_pb2.PredictionsRequest(model_name=model_name, input=input_data),
metadata=metadata,
)
try:
for resp in responses:
prediction = resp.prediction.decode("utf-8")
print(prediction)
except grpc.RpcError as e:
print(f"gRPC error: {e.details()}")
exit(1)
def infer_stream2(model_name, sequence_id, input_files, metadata):
response_queue = queue.Queue()
process_response_func = partial(
InferStream2.default_process_response, response_queue
)
client = InferStream2SimpleClient()
try:
client.start_stream(
model_name=model_name,
sequence_id=sequence_id,
process_response=process_response_func,
metadata=metadata,
)
sequence = input_files.split(",")
for input_file in sequence:
client.async_send_infer(input_file.strip())
for i in range(0, len(sequence)):
response = response_queue.get()
print(str(response))
print("Sequence completed!")
except grpc.RpcError as e:
print("infer_stream2 received error", e)
exit(1)
finally:
client.stop_stream()
client.stop()
def register(stub, model_name, mar_set_str, metadata):
mar_set = set()
if mar_set_str:
mar_set = set(mar_set_str.split(","))
marfile = f"{model_name}.mar"
print(f"## Check {marfile} in mar_set :", mar_set)
if marfile not in mar_set:
marfile = "https://torchserve.s3.amazonaws.com/mar_files/{}.mar".format(
model_name
)
print(f"## Register marfile: {marfile}\n")
params = {
"url": marfile,
"initial_workers": 1,
"synchronous": True,
"model_name": model_name,
}
try:
response = stub.RegisterModel(
management_pb2.RegisterModelRequest(**params), metadata=metadata
)
print(f"Model {model_name} registered successfully")
except grpc.RpcError as e:
print(f"Failed to register model {model_name}.")
print(str(e.details()))
exit(1)
def unregister(stub, model_name, metadata):
try:
response = stub.UnregisterModel(
management_pb2.UnregisterModelRequest(model_name=model_name),
metadata=metadata,
)
print(f"Model {model_name} unregistered successfully")
except grpc.RpcError as e:
print(f"Failed to unregister model {model_name}.")
print(str(e.details()))
exit(1)
if __name__ == "__main__":
parent_parser = argparse.ArgumentParser(add_help=False)
parent_parser.add_argument(
"model_name",
type=str,
default=None,
help="Name of the model used.",
)
parent_parser.add_argument(
"--auth-token",
dest="auth_token",
type=str,
default=None,
required=False,
help="Authorization token",
)
parser = argparse.ArgumentParser(
description="TorchServe gRPC client",
formatter_class=argparse.RawTextHelpFormatter,
)
subparsers = parser.add_subparsers(help="Action", dest="action")
infer_action_parser = subparsers.add_parser(
"infer", parents=[parent_parser], add_help=False
)
infer_stream_action_parser = subparsers.add_parser(
"infer_stream", parents=[parent_parser], add_help=False
)
infer_stream2_action_parser = subparsers.add_parser(
"infer_stream2", parents=[parent_parser], add_help=False
)
register_action_parser = subparsers.add_parser(
"register", parents=[parent_parser], add_help=False
)
unregister_action_parser = subparsers.add_parser(
"unregister", parents=[parent_parser], add_help=False
)
infer_action_parser.add_argument(
"model_input", type=str, default=None, help="Input for model for inference."
)
infer_stream_action_parser.add_argument(
"model_input",
type=str,
default=None,
help="Input for model for stream inference.",
)
infer_stream2_action_parser.add_argument(
"sequence_id",
type=str,
default=None,
help="Input for sequence id for stream inference.",
)
infer_stream2_action_parser.add_argument(
"input_files",
type=str,
default=None,
help="Comma separated list of input files",
)
register_action_parser.add_argument(
"mar_set",
type=str,
default=None,
nargs="?",
help="Comma separated list of mar models to be loaded using [model_name=]model_location format.",
)
args = parser.parse_args()
if args.auth_token:
metadata = (
("protocol", "gRPC"),
("session_id", "12345"),
("authorization", f"Bearer {args.auth_token}"),
)
else:
metadata = (("protocol", "gRPC"), ("session_id", "12345"))
if args.action == "infer":
infer(get_inference_stub(), args.model_name, args.model_input, metadata)
elif args.action == "infer_stream":
infer_stream(get_inference_stub(), args.model_name, args.model_input, metadata)
elif args.action == "infer_stream2":
infer_stream2(args.model_name, args.sequence_id, args.input_files, metadata)
elif args.action == "register":
register(get_management_stub(), args.model_name, args.mar_set, metadata)
elif args.action == "unregister":
unregister(get_management_stub(), args.model_name, metadata)
|
