update default sm num

ops/csrc/fp8/deep_gemm/jit_kernels/gemm.py

@@ -171,7 +171,7 @@ def auto_tuning_with_compilation(m, n, k, num_sms):
     return runtime, num_sms, smem_size
 
 
-def gemm_fp8_fp8_bf16_nt(lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, num_sms=112) -> None:
+def gemm_fp8_fp8_bf16_nt(lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, num_sms=132) -> None:
     """
     Do a normal GEMM with FP8 inputs and BF16 output, with 1x128 LHS scaling and 128x128 RHS scaling.
     LHS, RHS, RHS scaling factors, and output tensors must be in contiguous format.

ops/csrc/fp8/deep_gemm/jit_kernels/m_grouped_gemm.py

@@ -98,7 +98,7 @@ def auto_tuning_with_compilation_grouped_gemm_contiguous(m, n, k, num_groups, nu
 
 
 def m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(
-    lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, m_indices: Tensor, num_sms=112
+    lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, m_indices: Tensor, num_sms=132
 ) -> None:
     """
     Do a grouped GEMM (contiguous format) with FP8 inputs and BF16 output, with 1x128 LHS scaling and 128x128 RHS scaling.
@@ -215,7 +215,7 @@ def auto_tuning_with_compilation_grouped_gemm_masked(m, expected_m, n, k, num_gr
 
 
 def m_grouped_gemm_fp8_fp8_bf16_nt_masked(
-    lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, masked_m: Tensor, expected_m: int, num_sms=112
+    lhs: Tuple[Tensor, Tensor], rhs: Tuple[Tensor, Tensor], out: Tensor, masked_m: Tensor, expected_m: int, num_sms=132
 ) -> None:
     """
     Do a grouped GEMM (masked format) with FP8 inputs and BF16 output, with 1x128 LHS scaling and 128x128 RHS scaling.