Refactor JIT kernel and expert-pack directory layout (#36704)

test/registered/expert_pack/test_expert_pack_runtime.py

@@ -4,7 +4,6 @@ from __future__ import annotations
 
 import hashlib
 import json
-import sys
 import tempfile
 import unittest
 from dataclasses import replace
@@ -16,11 +15,11 @@ from sglang.test.ci.ci_register import register_cpu_ci
 
 register_cpu_ci(est_time=15, suite="base-a-test-cpu")
 
-ROOT = Path(__file__).resolve().parents[3]
-TOOLS = ROOT / "tools" / "expert_pack"
-sys.path.insert(0, str(TOOLS))
-
-from format import (  # noqa: E402
+from sglang.srt.layers.moe.expert_pack import (  # noqa: E402
+    ExpertPackStore,
+    _CacheSlot,
+)
+from sglang.srt.model_loader.expert_pack.format import (  # noqa: E402
     ENTRY_STRUCT,
     FLAG_IDENTITY_PAYLOAD,
     FLAG_TRIPLET_OBJECTS,
@@ -32,11 +31,6 @@ from format import (  # noqa: E402
     read_index,
 )
 
-from sglang.srt.layers.moe.expert_pack import (  # noqa: E402
-    ExpertPackStore,
-    _CacheSlot,
-)
-
 
 def _sha256(path: Path) -> str:
     return hashlib.sha256(path.read_bytes()).hexdigest()

test/registered/kernels/benchmark/bench_get_block_table.py

@@ -1,7 +1,7 @@
 import torch
 
 from sglang.kernels.jit.benchmark import marker
-from sglang.kernels.jit.minicpm_sala import get_block_table
+from sglang.kernels.ops.minicpm_sala import get_block_table
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(

test/registered/kernels/ops/test_get_block_table.py

@@ -1,7 +1,7 @@
 import pytest
 import torch
 
-from sglang.kernels.jit.minicpm_sala.get_block_table import get_block_table
+from sglang.kernels.ops.minicpm_sala.get_block_table import get_block_table
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=30, stage="base-b-kernel-unit", runner_config="1-gpu-large")