Merge pull request #50094 from tensorflow/geetachavan1-patch-4

Update ragged_tensor_to_variant_op.cc

tensorflow/core/kernels/ragged_tensor_to_variant_op.cc

@@ -176,7 +176,7 @@ class RaggedTensorToVariantOp : public OpKernel {
     // Unbatch the Ragged Tensor and encode the components.
     std::vector<RaggedTensor> ragged_components;
     auto batched_splits_top_vec =
-        batched_ragged_input.splits(0).vec<SPLIT_TYPE>();
+        batched_ragged_input.nested_splits[0].vec<SPLIT_TYPE>();
     int num_components = batched_splits_top_vec.size() - 1;
     OP_REQUIRES(context, num_components >= 0,
                 errors::Internal("Invalid split argument."));