@@ -183,7 +183,7 @@ def init_health_status(self) -> None:
183
183
name = "launched_expert_service_signal" ,
184
184
array = launched_expert_service_signal_data ,
185
185
dtype = np .int32 ,
186
- suffix = self .parallel_config .engine_worker_queue_port ,
186
+ suffix = self .parallel_config .engine_pid ,
187
187
create = False ,
188
188
)
189
189
while self .launched_expert_service_signal .value [self .local_rank % self .max_chips_per_node ] == 0 :
@@ -200,7 +200,7 @@ def init_health_status(self) -> None:
200
200
name = "worker_ready_signal" ,
201
201
array = workers_ready ,
202
202
dtype = np .int32 ,
203
- suffix = self .parallel_config .engine_worker_queue_port ,
203
+ suffix = self .parallel_config .engine_pid ,
204
204
create = False ,
205
205
)
206
206
self .worker_ready_signal .value [self .local_rank % self .max_chips_per_node ] = 1
@@ -418,7 +418,7 @@ def initialize_kv_cache(self) -> None:
418
418
name = "get_profile_block_num" ,
419
419
array = get_profile_block_num ,
420
420
dtype = np .int32 ,
421
- suffix = self .parallel_config .engine_worker_queue_port ,
421
+ suffix = self .parallel_config .engine_pid ,
422
422
create = False ,
423
423
)
424
424
self .get_profile_block_num_signal .value [0 ] = num_blocks_local
@@ -473,7 +473,7 @@ def load_model(self) -> None:
473
473
name = "loaded_model_signal" ,
474
474
array = loaded_model_signal_data ,
475
475
dtype = np .int32 ,
476
- suffix = self .parallel_config .engine_worker_queue_port ,
476
+ suffix = self .parallel_config .engine_pid ,
477
477
create = False ,
478
478
)
479
479
if self .ranks > 1 :
0 commit comments