apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[XPU] xpu support mm prefix cache (#5356)

Browse Source 
Co-authored-by: ddchenhao66 <dhaochen163.com>
This commit is contained in:
ddchenhao66
2025-12-03 19:07:34 +08:00
committed by GitHub
parent a4bb3e9960
commit 4e8096bd0d
3 changed files with 184 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
fastdeploy/config.py
View File

@@ -1694,9 +1694,9 @@ class FDConfig:
logger.info(
"Static Graph does not support to be started together with RL Training, and automatically switch to dynamic graph!"
)
if self.device_config is not None and self.device_config.device_type != "cuda":
if not current_platform.is_cuda():
self.graph_opt_config.use_cudagraph = False
logger.info(f"CUDAGraph only support on GPU, current device type is {self.device_config.device_type}!")
logger.info("CUDAGraph currently only support on GPU!")
if self.parallel_config.use_sequence_parallel_moe and self.graph_opt_config.use_cudagraph:
if self.scheduler_config.max_num_seqs < self.parallel_config.tensor_parallel_size:
self.parallel_config.use_sequence_parallel_moe = False