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] Support PaddleOCR-VL model for XPU (#4529)

Browse Source 
* [XPU] support PaddleOCR-VL in XPU

* [XPU] fix PaddleOCR-VL pos_emb_type
This commit is contained in:
Lucas
2025-10-28 20:35:04 +08:00
committed by GitHub
parent 2a9ed72533
commit 0a0c74e717
2 changed files with 109 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
fastdeploy/model_executor/layers/attention/xpu_attn_backend.py
View File

@@ -83,7 +83,9 @@ class XPUAttentionBackend(AttentionBackend):
self.rope_theta: float = (
10000.0 if fd_config.model_config.rope_theta is None else fd_config.model_config.rope_theta
)
self.rope_3d: bool = getattr(fd_config.model_config, "rope_3d", False)
self.rope_3d: bool = getattr(fd_config.model_config, "rope_3d", False) or getattr(
fd_config.model_config, "use_3d_rope", False
)
self.causal: bool = getattr(fd_config.model_config, "causal", True)
self.keep_pd_step_flag: bool = fd_config.speculative_config.model_type == "mtp"
self.rank: int = fd_config.parallel_config.tensor_parallel_rank