[Common] Disabled the tuned NVFP4 kernels (#2615)

* Disabled the tuned NVFP4 kernels
Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com>

* Disabled fast math in cpp tests
Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com>

---------
Signed-off-by: Oleg Goncharov <ogoncharov@nvidia.com>

tests/cpp/operator/test_cast_nvfp4_transpose.cu

@@ -677,11 +677,6 @@ std::vector<ActivationType> Activation_types = {
     ActivationType::Identity
 };
 
-std::vector<bool> use_fast_nvfp4_scaling_vec = {
-    false,
-    true
-};
-
 }  // namespace
 
 class FusedCastTransposeNVFP4TestSuite : public ::testing::TestWithParam
@@ -743,7 +738,7 @@ INSTANTIATE_TEST_SUITE_P(
         ::testing::ValuesIn(Activation_types),
         ::testing::ValuesIn(tensor_dims),
         ::testing::Values(DType::kBFloat16),
-        ::testing::ValuesIn(use_fast_nvfp4_scaling_vec)),
+        ::testing::Values(false)),
     [](const testing::TestParamInfo<FusedCastTransposeNVFP4TestSuite::ParamType>& info) {
         std::string name = to_string(std::get<0>(info.param));
       const auto& shape = std::get<1>(info.param);

transformer_engine/common/cast/nvfp4/quantize_transpose_nvfp4.cuh

@@ -1168,10 +1168,10 @@ void quantize_transpose(const Tensor &input, const Tensor *noop, Tensor *output,
   // TODO(Frank): Is there a better way to do this?
   bool return_transpose = output->has_columnwise_data();
 
-  if (!use_2d_quantization && (input.dtype() == DType::kBFloat16)) {
-    quantize_transpose_tuned_1D(input, noop, output, quant_config, stream);
-    return;
-  }
+  // if (!use_2d_quantization && (input.dtype() == DType::kBFloat16)) {
+  //   quantize_transpose_tuned_1D(input, noop, output, quant_config, stream);
+  //   return;
+  // }
 
   constexpr bool COMPUTE_ACTIVATIONS = false;
   using ParamOP = Empty;