Expose mxfp8 grouped gemm in torch ops

Summary: Expose mxfp8 grouped gemm by moving function declaration to torch_ops.h

Reviewed By: q10

Differential Revision: D81690096

Author: danielvegamyhre

fbgemm_gpu/experimental/gen_ai/src/quantize/include/fbgemm_gpu/torch_ops.h

@@ -27,6 +27,17 @@ at::Tensor f8f8bf16_rowwise_grouped_mm(
     std::optional<at::Tensor> offsets,
     at::Tensor& output);
 
+#else
+
+// Torch compliant MXFP8 grouped GEMM only on CUDA for now.
+at::Tensor mx8mx8bf16_grouped_mm(
+    at::Tensor XQ,
+    at::Tensor WQ,
+    at::Tensor x_scale,
+    at::Tensor w_scale,
+    at::Tensor offsets,
+    std::optional<at::Tensor> output = std::nullopt);
+
 #endif
 
 } // namespace fbgemm_gpu

fbgemm_gpu/experimental/gen_ai/src/quantize/quantize.cpp

@@ -64,13 +64,6 @@ at::Tensor f4f4bf16_grouped_stacked(
     std::optional<at::Tensor> global_scale = std::nullopt,
     std::optional<at::Tensor> starting_row_after_padding = std::nullopt,
     bool use_mx = true);
-at::Tensor mx8mx8bf16_grouped_mm(
-    at::Tensor XQ,
-    at::Tensor WQ,
-    at::Tensor x_scale,
-    at::Tensor w_scale,
-    at::Tensor offsets,
-    std::optional<at::Tensor> output = std::nullopt);
 at::Tensor f8f8bf16(
     at::Tensor XQ,
     at::Tensor WQ,