diff --git a/mlir/docs/Dialects/GPU.md b/mlir/docs/Dialects/GPU.md
index 8d4d2ca3e5743..511e4ef03ed60 100644
--- a/mlir/docs/Dialects/GPU.md
+++ b/mlir/docs/Dialects/GPU.md
@@ -121,7 +121,7 @@ func.func @main() {
     gpu.launch
         blocks(%0, %1, %2) in (%3 = %c1, %4 = %c1, %5 = %c1)
         threads(%6, %7, %8) in (%9 = %c2, %10 = %c1, %11 = %c1) {
-        gpu.printf "Hello from %d\n" %6 : index
+        gpu.printf "Hello from %d\n", %6 : index
         gpu.terminator
     }
     return
@@ -135,7 +135,7 @@ execution using `mlir-runner`. Alternatively, it can be translated into
 LLVM, expanding its utility within the system.
 
 ```
-mlir-opt example.mlir -gpu-lower-to-nvvm-pipeline = "cubin-chip=sm_90a cubin-features=+ptx80 opt-level=3"
+mlir-opt example.mlir -gpu-lower-to-nvvm-pipeline="cubin-chip=sm_90a cubin-features=+ptx80 opt-level=3"
 ```
 
 ### Module serialization