[Kernel] Classification cleanup: unify _jit_ naming, drop empty/model groups, add elementwise (RFC #29630) (#32148)

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

test/registered/kernels/ops/quantization/test_per_tensor_quant_fp8.py

@@ -6,7 +6,7 @@ import pytest
 import torch
 
 from sglang.kernels.jit.utils import get_ci_test_range
-from sglang.kernels.ops.quantization._jit_per_tensor_quant_fp8 import (
+from sglang.kernels.ops.quantization.per_tensor_quant_fp8 import (
     per_tensor_quant_fp8,
 )
 from sglang.test.ci.ci_register import register_cuda_ci

test/registered/kernels/ops/quantization/test_per_token_group_quant.py

@@ -22,14 +22,14 @@ import pytest
 import torch
 
 from sglang.kernels.jit.utils import get_ci_test_range
-from sglang.kernels.ops.quantization._jit_per_token_group_quant import (
-    per_token_group_quant,
-)
 from sglang.kernels.ops.quantization.fp8_kernel import (
     create_per_token_group_quant_fp8_output_scale,
     fp8_dtype,
     fp8_max,
 )
+from sglang.kernels.ops.quantization.per_token_group_quant import (
+    per_token_group_quant,
+)
 from sglang.test.ci.ci_register import register_cuda_ci
 
 register_cuda_ci(est_time=90, stage="base-b-kernel-unit", runner_config="1-gpu-large")

test/registered/kernels/ops/quantization/test_per_token_group_quant_8bit_v2.py

@@ -4,7 +4,7 @@ import pytest
 import torch
 
 from sglang.kernels.jit.utils import get_ci_test_range
-from sglang.kernels.ops.quantization._jit_per_token_group_quant_8bit_v2 import (
+from sglang.kernels.ops.quantization.per_token_group_quant_8bit_v2 import (
     per_token_group_quant_8bit_v2,
 )
 from sglang.test.ci.ci_register import register_cuda_ci