Cannot perform quantization-aware-finetuning due to NaN values

ValueError is raised as a result of trying to cast NaN values to integers.
See PR: https://github.com/kssteven418/I-BERT/pull/30.