apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-08 10:00:29 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[XPU] support XPU VL model inference (#4030)

Browse Source 
* [XPU] support XPU VL model inference

* fix image op import and device check

* rebase develop

* fix perf
This commit is contained in:
Lucas
2025-09-25 14:34:15 +08:00
committed by GitHub
parent e36eccfdad
commit 87179cb744
18 changed files with 1300 additions and 146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
fastdeploy/engine/sched/resource_manager_v1.py
View File

@@ -30,6 +30,7 @@ import paddle
from fastdeploy.engine.request import Request, RequestOutput, RequestStatus, RequestType
from fastdeploy.engine.resource_manager import ResourceManager
from fastdeploy.metrics.metrics import main_process_metrics
from fastdeploy.platforms import current_platform
from fastdeploy.utils import llm_logger
@@ -157,6 +158,7 @@ class ResourceManagerV1(ResourceManager):
# TODO: set condition to new _get_num_new_tokens
num_new_tokens = request.need_prefill_tokens - request.num_computed_tokens
num_new_tokens = min(num_new_tokens, token_budget)
request.with_image = False
if not self.config.model_config.enable_mm:
return num_new_tokens
@@ -219,7 +221,10 @@ class ResourceManagerV1(ResourceManager):
grid_thw.extend([[2, one[1], one[2]]] * (one[0] // 2))
grid_thw = paddle.to_tensor(grid_thw, dtype="int64")
from fastdeploy.model_executor.ops.gpu import get_img_boundaries
if current_platform.is_xpu():
from fastdeploy.model_executor.ops.xpu import get_img_boundaries
else:
from fastdeploy.model_executor.ops.gpu import 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