Fix f8f8bf16 function signature when cuda disabled (#2605)

Summary: Pull Request resolved: #2605 Since D57442310 changed f8f8bf16 function signature, changing the function signature back fixes the issue Reviewed By: yuguo68 Differential Revision: D57529592 fbshipit-source-id: afabd1c8de60d4797dd1688863bea0650f14b8fa
pytorch · May 19, 2024 · 06e78e1 · 06e78e1
1 parent 79aba2c
commit 06e78e1
Showing 1 changed file with 8 additions and 0 deletions.
diff --git a/fbgemm_gpu/experimental/gen_ai/src/quantize/cutlass_extensions.cu b/fbgemm_gpu/experimental/gen_ai/src/quantize/cutlass_extensions.cu
@@ -1896,6 +1896,14 @@ at::Tensor f8f8bf16_cublas(
       "CUDA version is older than 12.0"); // requires CUDA>=12
 }
 at::Tensor f8f8bf16(
+    at::Tensor XQ, // FP8
+    at::Tensor WQ, // FP8
+    at::Tensor scale,
+    bool use_fast_accum) {
+  throw std::runtime_error(
+      "CUDA version is older than 12.0"); // requires CUDA>=12
+}
+at::Tensor f8f8bf16_tensorwise(
     at::Tensor XQ, // FP8
     at::Tensor WQ, // FP8
     double scale,