apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 08:21:53 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

test_abort (#6743)

Browse Source
This commit is contained in:
qwes5s5
2026-03-17 14:06:40 +08:00
committed by GitHub
parent eab429d05e
commit 3b7507a4c2
11 changed files with 132 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/engine/common_engine.py
+1 -14
View File
@@ -786,7 +786,6 @@ class EngineService:
max_num_batched_tokens=self.cfg.scheduler_config.max_num_batched_tokens,
batch=num_prefill_batch,
)
tasks = [task for task in tasks if task.request_id not in self.resource_manager.abort_req_ids_set]
for task in tasks:
task.metrics.engine_get_req_time = time.time()
trace_print(LoggingEventName.REQUEST_QUEUE_END, task.request_id, getattr(task, "user", ""))
@@ -851,7 +850,6 @@ class EngineService:
max_num_batched_tokens=max_num_batched_tokens,
batch=num_prefill_batch,
)
tasks = [task for task in tasks if task.request_id not in self.resource_manager.abort_req_ids_set]
for task in tasks:
task.metrics.engine_get_req_time = time.time()
trace_print(LoggingEventName.REQUEST_QUEUE_END, task.request_id, getattr(task, "user", ""))
@@ -1178,19 +1176,8 @@ class EngineService:
if status_value is not None and status_value == RequestStatus.ABORT.value:
req_id = data["request_id"]
self.llm_logger.info(f"Receive abort request, req_id: {req_id}")
self.resource_manager.abort_req_ids_set.add(req_id)
if envs.ENABLE_V1_KVCACHE_SCHEDULER:
if req_id in self.resource_manager.requests:
req = self.resource_manager.requests[req_id]
task = self.resource_manager._prepare_preempt_task(req)
self.engine_worker_queue.put_tasks(([task], self.resource_manager.real_bsz))
self.llm_logger.info(f"put abort task in engine worker queue, req_id: {req_id}")
else:
self.scheduler._recycle(req_id)
self.llm_logger.info(
f"req_id:{req_id} has not been allocated any resources, recycled it in scheduler"
)
self.resource_manager.abort_req_ids_set.remove(req_id)
self.resource_manager.add_abort_req_ids(req_id)
continue
err_msg = None
try: