diff --git a/cpp/include/nvtext/wordpiece_tokenize.hpp b/cpp/include/nvtext/wordpiece_tokenize.hpp
index 85c0f1406bd..025cae6bf54 100644
--- a/cpp/include/nvtext/wordpiece_tokenize.hpp
+++ b/cpp/include/nvtext/wordpiece_tokenize.hpp
@@ -66,7 +66,7 @@ struct wordpiece_vocabulary {
  * @param input Strings for the vocabulary
  * @param stream CUDA stream used for device memory operations and kernel launches
  * @param mr Device memory resource used to allocate the returned column's device memory
- * @return Object to be used with nvtext::tokenize_with_vocabulary
+ * @return Object to be used with nvtext::wordpiece_tokenize
  */
 std::unique_ptr<wordpiece_vocabulary> load_wordpiece_vocabulary(
   cudf::strings_column_view const& input,
@@ -103,6 +103,8 @@ std::unique_ptr<wordpiece_vocabulary> load_wordpiece_vocabulary(
  *
  * Any null row entry results in a corresponding null entry in the output.
  *
+ * @throw std::invalid_argument If `max_words_per_row` is less than 0.
+ *
  * @param input Strings column to tokenize
  * @param vocabulary Used to lookup tokens within `input`
  * @param max_words_per_row Maximum number of words to tokenize for each row.
diff --git a/cpp/src/text/wordpiece_tokenize.cu b/cpp/src/text/wordpiece_tokenize.cu
index c21200e5c5b..fb44d671902 100644
--- a/cpp/src/text/wordpiece_tokenize.cu
+++ b/cpp/src/text/wordpiece_tokenize.cu
@@ -434,7 +434,6 @@ __device__ cudf::size_type wp_tokenize_fn(cudf::string_view word,
 template <typename MapRefType, typename SubMapRefType>
 CUDF_KERNEL void tokenize_all_kernel(cudf::device_span<int64_t const> d_edges,
                                      char const* d_chars,
-                                     // int64_t offset,
                                      MapRefType const d_map,
                                      SubMapRefType const d_sub_map,
                                      cudf::size_type unk_id,
@@ -447,7 +446,7 @@ CUDF_KERNEL void tokenize_all_kernel(cudf::device_span<int64_t const> d_edges,
   auto const word_end = thrust::find(thrust::seq, begin, end, ' ');
   auto const size     = static_cast<cudf::size_type>(cuda::std::distance(begin, word_end));
   if (size == 0) { return; }
-  auto d_output = d_tokens + d_edges[idx];  // - offset;
+  auto d_output = d_tokens + d_edges[idx];
   if (size >= max_word_size) {
     *d_output = unk_id;
     return;
@@ -833,7 +832,8 @@ std::unique_ptr<cudf::column> wordpiece_tokenize(cudf::strings_column_view const
                                                  rmm::cuda_stream_view stream,
                                                  rmm::device_async_resource_ref mr)
 {
-  CUDF_EXPECTS(max_words_per_row >= 0, "Invalid value for max_words_per_row argument");
+  CUDF_EXPECTS(
+    max_words_per_row >= 0, "Invalid value for max_words_per_row argument", std::invalid_argument);
 
   auto const output_type = cudf::data_type{cudf::type_to_id<cudf::size_type>()};
   if (input.size() == input.null_count()) {
diff --git a/cpp/tests/text/subword_tests.cpp b/cpp/tests/text/subword_tests.cpp
index b01042abf61..ca02f17672e 100644
--- a/cpp/tests/text/subword_tests.cpp
+++ b/cpp/tests/text/subword_tests.cpp
@@ -170,4 +170,10 @@ TEST(TextSubwordTest, WordPieceErrors)
   auto nulls = cudf::test::strings_column_wrapper({"", "", ""}, {false, false, false});
   EXPECT_THROW(nvtext::load_wordpiece_vocabulary(cudf::strings_column_view(nulls)),
                std::invalid_argument);
+
+  auto vocabulary = cudf::test::strings_column_wrapper({"x"});
+  auto vocab      = nvtext::load_wordpiece_vocabulary(cudf::strings_column_view(vocabulary));
+  auto input      = cudf::test::strings_column_wrapper({"  "});
+  EXPECT_THROW(nvtext::wordpiece_tokenize(cudf::strings_column_view(input), *vocab, -1),
+               std::invalid_argument);
 }