From d5f77f14d0eb3b8dbcf968eae443d6e5a1c79f4d Mon Sep 17 00:00:00 2001
From: shiyuan680 <72335504+shiyuan680@users.noreply.github.com>
Date: Wed, 26 Nov 2025 11:06:12 +0800
Subject: [PATCH] mkdir triton package and move triton files (#4420)

### What this PR does / why we need it?
mkdir triton package and move triton files

- vLLM version: v0.11.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/2918c1b49c88c29783c86f78d2c4221cb9622379

Signed-off-by: shiyuan680 <917935075@qq.com>
---
 vllm_ascend/ops/triton/__init__.py                  | 0
 vllm_ascend/ops/triton/fla/__init__.py              | 0
 vllm_ascend/ops/{ => triton/fla}/fla.py             | 0
 vllm_ascend/ops/{ => triton/fla}/sigmoid_gating.py  | 0
 vllm_ascend/ops/triton/mamba/__init__.py            | 0
 vllm_ascend/ops/{ => triton/mamba}/casual_conv1d.py | 0
 vllm_ascend/patch/worker/patch_triton.py            | 9 +++++----
 7 files changed, 5 insertions(+), 4 deletions(-)
 create mode 100644 vllm_ascend/ops/triton/__init__.py
 create mode 100644 vllm_ascend/ops/triton/fla/__init__.py
 rename vllm_ascend/ops/{ => triton/fla}/fla.py (100%)
 rename vllm_ascend/ops/{ => triton/fla}/sigmoid_gating.py (100%)
 create mode 100644 vllm_ascend/ops/triton/mamba/__init__.py
 rename vllm_ascend/ops/{ => triton/mamba}/casual_conv1d.py (100%)

diff --git a/vllm_ascend/ops/triton/__init__.py b/vllm_ascend/ops/triton/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/vllm_ascend/ops/triton/fla/__init__.py b/vllm_ascend/ops/triton/fla/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/vllm_ascend/ops/fla.py b/vllm_ascend/ops/triton/fla/fla.py
similarity index 100%
rename from vllm_ascend/ops/fla.py
rename to vllm_ascend/ops/triton/fla/fla.py
diff --git a/vllm_ascend/ops/sigmoid_gating.py b/vllm_ascend/ops/triton/fla/sigmoid_gating.py
similarity index 100%
rename from vllm_ascend/ops/sigmoid_gating.py
rename to vllm_ascend/ops/triton/fla/sigmoid_gating.py
diff --git a/vllm_ascend/ops/triton/mamba/__init__.py b/vllm_ascend/ops/triton/mamba/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/vllm_ascend/ops/casual_conv1d.py b/vllm_ascend/ops/triton/mamba/casual_conv1d.py
similarity index 100%
rename from vllm_ascend/ops/casual_conv1d.py
rename to vllm_ascend/ops/triton/mamba/casual_conv1d.py
diff --git a/vllm_ascend/patch/worker/patch_triton.py b/vllm_ascend/patch/worker/patch_triton.py
index cc550ccc..eb3f300b 100644
--- a/vllm_ascend/patch/worker/patch_triton.py
+++ b/vllm_ascend/patch/worker/patch_triton.py
@@ -3,11 +3,12 @@ import vllm.model_executor.layers.fla.ops.fused_recurrent
 import vllm.model_executor.layers.fla.ops.layernorm_guard
 import vllm.model_executor.layers.mamba.ops.causal_conv1d
 
-from vllm_ascend.ops.casual_conv1d import (causal_conv1d_fn,
-                                           causal_conv1d_update_npu)
-from vllm_ascend.ops.fla import LayerNormFn, torch_chunk_gated_delta_rule
-from vllm_ascend.ops.sigmoid_gating import \
+from vllm_ascend.ops.triton.fla.fla import (LayerNormFn,
+                                            torch_chunk_gated_delta_rule)
+from vllm_ascend.ops.triton.fla.sigmoid_gating import \
     fused_recurrent_gated_delta_rule_fwd_kernel
+from vllm_ascend.ops.triton.mamba.casual_conv1d import (
+    causal_conv1d_fn, causal_conv1d_update_npu)
 
 vllm.model_executor.layers.mamba.ops.causal_conv1d.causal_conv1d_update = causal_conv1d_update_npu
 vllm.model_executor.layers.mamba.ops.causal_conv1d.causal_conv1d_fn = causal_conv1d_fn