[Kernel] Migrate DSA + DSV4 attention kernels to sglang.kernels (RFC #29630, Phase 2.5, 5/7) (#30792)

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

test/manual/layers/attention/dsa/test_act_quant_triton.py

@@ -9,8 +9,8 @@ from typing import Tuple
 import pytest
 import torch
 
-from sglang.srt.layers.attention.dsa.tilelang_kernel import act_quant
-from sglang.srt.layers.attention.dsa.triton_kernel import act_quant as act_quant_triton
+from sglang.kernels.ops.attention.dsa.tilelang_kernel import act_quant
+from sglang.kernels.ops.attention.dsa.triton_kernel import act_quant as act_quant_triton
 
 
 def benchmark_kernel(

test/manual/layers/attention/dsa/test_get_k_scale_triton_kernel.py

@@ -1,6 +1,6 @@
 import torch
 
-from sglang.srt.layers.attention.dsa.index_buf_accessor import (
+from sglang.kernels.ops.attention.dsa.index_buf_accessor import (
     _get_k_and_s_triton_kernel,
 )
 

test/manual/layers/attention/dsa/test_index_buf_accessor.py

@@ -13,7 +13,7 @@ Test coverage:
 import pytest
 import torch
 
-from sglang.srt.layers.attention.dsa.index_buf_accessor import GetK, GetKAndS, GetS
+from sglang.kernels.ops.attention.dsa.index_buf_accessor import GetK, GetKAndS, GetS
 
 
 class MockDSATokenToKVPool:

test/registered/jit/deepseek_v4/test_fp4_indexer.py

@@ -15,7 +15,7 @@ from sglang.kernels.ops.attention.deepseek_v4_rope import (
     apply_rotary_emb_triton,
     precompute_freqs_cis,
 )
-from sglang.srt.layers.attention.dsv4.fp4_indexer import (
+from sglang.kernels.ops.attention.dsv4.fp4_indexer import (
     quantize_fp4_indexer_tensor,
     store_fp4_index_k_cache,
 )

test/registered/jit/test_fused_store_index_cache.py

@@ -188,7 +188,7 @@ def _reference_quantize_and_store(
 
 def _import_act_quant():
     try:
-        from sglang.srt.layers.attention.dsa.triton_kernel import act_quant
+        from sglang.kernels.ops.attention.dsa.triton_kernel import act_quant
 
         return act_quant
     except Exception:

test/registered/kernels/test_dsa_indexer.py

@@ -720,7 +720,7 @@ class TestDSAIndexer(CustomTestCase):
         self.assertEqual(indexer.layer_id, self.config["layer_id"])
 
     @patch("sglang.srt.layers.attention.dsa.dsa_indexer.deep_gemm")
-    @patch("sglang.srt.layers.attention.dsa.triton_kernel.act_quant")
+    @patch("sglang.kernels.ops.attention.dsa.triton_kernel.act_quant")
     def test_forward_extend_mode(self, mock_act_quant, mock_deep_gemm):
         """Test indexer forward pass in extend mode."""
         if not self.supports_fp8:
@@ -802,7 +802,7 @@ class TestDSAIndexer(CustomTestCase):
         )
 
     @patch("sglang.srt.layers.attention.dsa.dsa_indexer.deep_gemm")
-    @patch("sglang.srt.layers.attention.dsa.triton_kernel.act_quant")
+    @patch("sglang.kernels.ops.attention.dsa.triton_kernel.act_quant")
     def test_forward_decode_mode(self, mock_act_quant, mock_deep_gemm):
         """Test indexer forward pass in decode mode."""
         if not self.supports_fp8:

test/registered/kernels/test_dsa_transform_index.py

@@ -3,8 +3,8 @@ from unittest.mock import patch
 
 import torch
 
-import sglang.srt.layers.attention.dsa.transform_index as transform_index_module
-from sglang.srt.layers.attention.dsa.transform_index import (
+import sglang.kernels.ops.attention.dsa.transform_index as transform_index_module
+from sglang.kernels.ops.attention.dsa.transform_index import (
     transform_index_page_table_decode_fast,
     transform_index_page_table_prefill_fast,
 )