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

[Ops][Misc] Refactor and optimize CausalConv1d for Ascend (#7495)

Browse Source 
### What this PR does / why we need it?
During the prefill phase of Qwen3-Next and Qwen3.5, the
`torch.ops._C_ascend.causal_conv1d_fn` operator exhibits significant
performance bottlenecks. To address this, we have re-implemented the
optimization using `torch.ops._C_ascend.npu_causal_conv1d_custom`.

### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
1 accuracy test
```
[2026-03-20 16:44:22,961] [ais_bench] [INFO] Start launch task state board ...
+-----------------------------+-----------+------------+-------------+----------+-------------------------------------------+---------------------+
| Task Name                   |   Process | Progress   | Time Cost   | Status   | Log Path                                  | Extend Parameters   |
+=============================+===========+============+=============+==========+===========================================+=====================+
| vllm-api-general-chat/gsm8k |   2918978 | NA         | 0:00:01     | finish   | logs/eval/vllm-api-general-chat/gsm8k.out | None                |
+-----------------------------+-----------+------------+-------------+----------+-------------------------------------------+---------------------+
[2026-03-20 16:44:34,284] [ais_bench] [INFO] Evaluation tasks completed.
[2026-03-20 16:44:34,287] [ais_bench] [INFO] Summarizing evaluation results...
dataset    version    metric    mode      vllm-api-general-chat
---------  ---------  --------  ------  -----------------------
gsm8k      271d0b     accuracy  gen                       96.21
```
2 ut modify test
`pytest -sv
/home/c30006096/vllm-ascend/tests/e2e/nightly/single_node/ops/singlecard_ops/triton/test_causal_conv1d.py::test_ascend_causal_conv1d`

- vLLM version: v0.17.0
- vLLM main:
8b6325758c

Signed-off-by: wenba0 <3054239545@qq.com>
Signed-off-by: jiaojiao <56385650+wenba0@users.noreply.github.com>
This commit is contained in:
jiaojiao
2026-03-24 00:07:12 +08:00
committed by GitHub
parent e942b62d74
commit 1de805ce0a
16 changed files with 907 additions and 554 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
csrc/causal_conv1d/op_host/causal_conv1d_infershape.cpp
View File

@@ -14,7 +14,7 @@
* \brief
*/
#include "register/op_impl_registry.h"
#include "error_log.h"
#include "log/log.h"
using namespace ge;
@@ -23,27 +23,19 @@ static constexpr int64_t IDX_0 = 0;
static ge::graphStatus InferShapeCausalConv1d(gert::InferShapeContext* context)
{
// OPS_LOG_D(context->GetNodeName(), "Begin to do InferShapeCausalConv1d");
OP_LOGD(context->GetNodeName(), "Begin to do InferShapeCausalConv1d");
// get input shapes
const gert::Shape* xShape = context->GetInputShape(IDX_0);
OP_CHECK_NULL_WITH_CONTEXT(context, xShape);
// get output shapes
gert::Shape* yShape = context->GetOutputShape(IDX_0);
OP_CHECK_NULL_WITH_CONTEXT(context, yShape);
// 填充输出shape大小
auto xShapeSize = xShape->GetDimNum();
yShape->SetDimNum(xShapeSize);
for (size_t i = 0; i < xShapeSize; i++) {
int64_t dim = xShape->GetDim(i);
yShape->SetDim(i, dim);
}
*yShape = *xShape;
// OPS_LOG_D(context->GetNodeName(), "End to do InferShapeCausalConv1d");
OP_LOGD(context->GetNodeName(), "End to do InferShapeCausalConv1d");
return GRAPH_SUCCESS;
}
IMPL_OP_INFERSHAPE(CausalConv1d).InferShape(InferShapeCausalConv1d);
} // namespace ops
} // namespace ops