apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-30 11:26:39 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

[Feature] Set v1 scheduler as default in develop (#3807)

Browse Source 
* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* Set scheduler v1 as default

* close V1 in guided_decoding

* fix vl ci

* close V1 in guided_decoding
This commit is contained in:
chenjian
2025-09-04 15:16:56 +08:00
committed by GitHub
parent e83251699f
commit 22c165d6dd
8 changed files with 50 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
fastdeploy/engine/sched/resource_manager_v1.py
View File

@@ -137,6 +137,8 @@ class ResourceManagerV1(ResourceManager):
self.to_be_rescheduled_request_id_set.add(preempted_req.request_id)
preempted_reqs.append(preempted_req)
scheduled_reqs.append(self._prepare_preempt_task(preempted_req))
main_process_metrics.num_requests_waiting.inc(1)
main_process_metrics.num_requests_running.dec(1)
if preempted_req == request:
# No more request to preempt.
can_schedule = False
@@ -155,6 +157,7 @@ class ResourceManagerV1(ResourceManager):
if not self.config.model_config.enable_mm:
return num_new_tokens
request.with_image = False
inputs = request.multimodal_inputs
if inputs.get("patch_idx", None) is not None and inputs.get("patch_map", None) is not None:
pre_end_idx = request.num_computed_tokens
@@ -198,8 +201,6 @@ class ResourceManagerV1(ResourceManager):
and inputs.get("image_patch_id", None) is not None
and inputs.get("grid_thw", None) is not None
):
request.with_image = False
input_ids_lst = request.prompt_token_ids + request.output_token_ids
input_ids = paddle.to_tensor(input_ids_lst, dtype="int64")
input_ids = paddle.to_tensor(input_ids_lst, dtype="int64")
@@ -383,6 +384,8 @@ class ResourceManagerV1(ResourceManager):
token_budget -= num_new_tokens
request.num_computed_tokens += num_new_tokens
request.status = RequestStatus.RUNNING
main_process_metrics.num_requests_waiting.dec(1)
main_process_metrics.num_requests_running.inc(1)
allocated_position = self.get_available_position()
request.idx = allocated_position
self.tasks_list[allocated_position] = request
@@ -413,6 +416,8 @@ class ResourceManagerV1(ResourceManager):
token_budget -= num_new_tokens
request.num_computed_tokens += num_new_tokens
request.status = RequestStatus.RUNNING
main_process_metrics.num_requests_waiting.dec(1)
main_process_metrics.num_requests_running.inc(1)
else:
if self.config.cache_config.enable_prefix_caching:
self._free_blocks(request)