[Kernel Slimming] Remove sgl-kernel AOT marlin kernels (#19241)

sgl-kernel/python/sgl_kernel/gemm.py

@@ -1,7 +1,6 @@
 from typing import Optional, Tuple
 
 import torch
-from sgl_kernel.scalar_type import ScalarType
 from sgl_kernel.utils import _get_cache_buf
 
 
@@ -506,46 +505,6 @@ def scaled_fp4_experts_quant(
 
 
 # GPTQ kernels
-def gptq_marlin_gemm(
-    a: torch.Tensor,
-    c: Optional[torch.Tensor],
-    b_q_weight: torch.Tensor,
-    b_scales: torch.Tensor,
-    global_scale: Optional[torch.Tensor],
-    b_zeros: Optional[torch.Tensor],
-    g_idx: Optional[torch.Tensor],
-    perm: Optional[torch.Tensor],
-    workspace: torch.Tensor,
-    b_q_type: ScalarType,
-    size_m: int,
-    size_n: int,
-    size_k: int,
-    is_k_full: bool = True,
-    use_atomic_add: bool = False,
-    use_fp32_reduce: bool = False,
-    is_zp_float: bool = False,
-) -> torch.Tensor:
-    return torch.ops.sgl_kernel.gptq_marlin_gemm(
-        a,
-        c,
-        b_q_weight,
-        b_scales,
-        global_scale,
-        b_zeros,
-        g_idx,
-        perm,
-        workspace,
-        b_q_type.id,
-        size_m,
-        size_n,
-        size_k,
-        is_k_full,
-        use_atomic_add,
-        use_fp32_reduce,
-        is_zp_float,
-    )
-
-
 def gptq_gemm(
     a: torch.Tensor,
     b_q_weight: torch.Tensor,