EngineX/xc-llm-kunlun
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Feature] support deepseek v3/r1/v3.2 (#78)

Browse Source 
* [Feature] support deepseek v3/r1/v3.2

* fix gpt_oss

* update readme

* update readme

---------

Co-authored-by: hanhaowen <hanhaowen@baidu.com>
This commit is contained in:
baoqian426
2026-01-05 22:55:35 +08:00
committed by GitHub
parent 07bc24a555
commit ee0f50e68f
27 changed files with 5760 additions and 621 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm_kunlun/ops/activation.py
View File

@@ -108,7 +108,7 @@ class SiluAndMul(CustomOp):
d = x.shape[-1] // 2
output_shape = (x.shape[:-1] + (d, ))
out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
xtorch_ops.swiglu(x, out)
torch.ops._C.silu_and_mul(out, x)
return out
def forward_xpu(self, x: torch.Tensor) -> torch.Tensor: