@@ -59,9 +59,7 @@ def execute(self, requests):
59
59
def _execute_processed_requests (self , processed_requests ):
60
60
responses = []
61
61
for processed_request in processed_requests :
62
- response = pb_utils .InferenceResponse (
63
- error = pb_utils .TritonError (message = "not cancelled" )
64
- )
62
+ error = pb_utils .TritonError (message = "not cancelled" )
65
63
object_to_check_cancelled = None
66
64
if "response_sender" in processed_request :
67
65
object_to_check_cancelled = processed_request ["response_sender" ]
@@ -78,14 +76,16 @@ def _execute_processed_requests(self, processed_requests):
78
76
+ str (time_elapsed )
79
77
+ " s"
80
78
)
81
- response = None
79
+ error = pb_utils .TritonError (
80
+ message = "cancelled" , code = pb_utils .TritonError .CANCELLED
81
+ )
82
82
break
83
83
self ._logger .log_info (
84
84
"[execute_cancel] Request not cancelled at "
85
85
+ str (time_elapsed )
86
86
+ " s"
87
87
)
88
- responses .append (response )
88
+ responses .append (pb_utils . InferenceResponse ( error = error ) )
89
89
return responses
90
90
91
91
def _execute_decoupled (self , processed_requests ):
@@ -94,9 +94,7 @@ def response_thread(execute_processed_requests, processed_requests):
94
94
responses = execute_processed_requests (processed_requests )
95
95
for i in range (len (responses )): # len(responses) == len(processed_requests)
96
96
response_sender = processed_requests [i ]["response_sender" ]
97
- response = responses [i ]
98
- if response != None :
99
- response_sender .send (response )
97
+ response_sender .send (responses [i ])
100
98
response_sender .send (
101
99
flags = pb_utils .TRITONSERVER_RESPONSE_COMPLETE_FINAL
102
100
)
0 commit comments