apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-11 11:30:20 +08:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

[Bug Fix] Fix V1 video bug (#3387)

Browse Source
This commit is contained in:
ming1753
2025-08-14 09:49:22 +08:00
committed by GitHub
parent 192f9caab4
commit a375378cc1
1 changed files with 25 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
fastdeploy/engine/sched/resource_manager_v1.py
View File

@@ -142,6 +142,10 @@ class ResourceManagerV1(ResourceManager):
input_ids_lst = request.prompt_token_ids + request.output_token_ids 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")
input_ids = paddle.to_tensor(input_ids_lst, dtype="int64")
image_patch_id = inputs["image_patch_id"]
if request.multimodal_img_boundaries is None:
grid_thw = [] grid_thw = []
for one in inputs["grid_thw"]: for one in inputs["grid_thw"]:
if one[0] == 1: if one[0] == 1:
@@ -149,19 +153,20 @@ class ResourceManagerV1(ResourceManager):
else: else:
grid_thw.extend([[2, one[1], one[2]]] * (one[0] // 2)) grid_thw.extend([[2, one[1], one[2]]] * (one[0] // 2))
image_patch_id = inputs["image_patch_id"]
grid_thw = paddle.to_tensor(grid_thw, dtype="int64") grid_thw = paddle.to_tensor(grid_thw, dtype="int64")
if request.multimodal_img_boundaries is None:
from fastdeploy.model_executor.ops.gpu import get_img_boundaries from fastdeploy.model_executor.ops.gpu import get_img_boundaries
request.multimodal_img_boundaries = get_img_boundaries( request.multimodal_img_boundaries = get_img_boundaries(
task_input_ids=input_ids, grid_thw=grid_thw, image_patch_id=image_patch_id task_input_ids=input_ids, grid_thw=grid_thw, image_patch_id=image_patch_id
).numpy() ).numpy()
grid_thw = grid_thw.numpy().reshape([-1, 3])
inputs["grid_thw"] = grid_thw
grid_thw = inputs["grid_thw"]
img_boundaries_idx = request.multimodal_img_boundaries[0] img_boundaries_idx = request.multimodal_img_boundaries[0]
img_num_per_boundary = request.multimodal_img_boundaries[1] img_num_per_boundary = request.multimodal_img_boundaries[1]
ori_prompt_len = img_boundaries_idx[-1].item() ori_prompt_len = img_boundaries_idx[-1].item()
grid_thw = grid_thw.numpy().reshape([-1, 3])
pre_end_idx = request.num_computed_tokens pre_end_idx = request.num_computed_tokens
new_end_idx = pre_end_idx + num_new_tokens new_end_idx = pre_end_idx + num_new_tokens
if new_end_idx < ori_prompt_len and input_ids[new_end_idx - 1] == image_patch_id: if new_end_idx < ori_prompt_len and input_ids[new_end_idx - 1] == image_patch_id:
@@ -421,9 +426,15 @@ class ResourceManagerV1(ResourceManager):
self.running.remove(request) self.running.remove(request)
request.status = RequestStatus.FINISHED request.status = RequestStatus.FINISHED
self._free_blocks(request) self._free_blocks(request)
if request.request_id in self.to_be_rescheduled_request_id_set: # finished after preempted, blocks have been recycled. if (
self.to_be_rescheduled_request_id_set.remove(request.request_id) # just remove from to_be_rescheduled_request_id_set request.request_id in self.to_be_rescheduled_request_id_set
if request in self.waiting: # after finished, this request still scheduled from preempted to waiting, unexpected error, should not be here ): # finished after preempted, blocks have been recycled.
self.to_be_rescheduled_request_id_set.remove(
request.request_id
) # just remove from to_be_rescheduled_request_id_set
if (
request in self.waiting
): # after finished, this request still scheduled from preempted to waiting, unexpected error, should not be here
raise RuntimeError(f"request {request.request_id} scheduled into waiting list, after finished") raise RuntimeError(f"request {request.request_id} scheduled into waiting list, after finished")
self.tasks_list[request.idx] = None self.tasks_list[request.idx] = None
self.stop_flags[request.idx] = True self.stop_flags[request.idx] = True