Minor fixes (microsoft#18949)

### Description
<!-- Describe your changes. -->



### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cpu/torch_interop_utils/custom_function_fw.cc

@@ -181,7 +181,6 @@ py::object finalize_training_mode_forward(
   }
 
   if (kernel_info.is_first_run) {
-    std::cout << "666666666666666666666666.  py_fn->materialize_grads:" << py_fn->materialize_grads << std::endl;
     get_materialize_grads_once(forward_output_tensors, py_fn->materialize_grads, kernel_info);
 
     if (kernel_info.safe_run_enabled) {

orttraining/orttraining/python/training/utils/hooks/_statistics_subscriber.py

@@ -252,7 +252,10 @@ def _summarize_tensor(
             min_buckets[i] = bucket.min()
             max_buckets[i] = bucket.max()
             mean_buckets[i] = bucket.sum()
-            std_buckets[i] = bucket.std()
+
+            # Only calculate std for float types, otherwise it will throw exception.
+            if bucket.dtype in [torch.float16, torch.float32, torch.float64]:
+                std_buckets[i] = bucket.std()
 
         # Reduction across all buckets
         num_nan = nan_buckets.sum()