Getting 0 accuracy by using pre-trained model

I am getting 0 accuracy when I am using the `pre-trained model` and then testing this model on the IIIT-5K dataset (converted to LMDB format by using [CRNN Create LMDB format](https://github.com/clovaai/deep-text-recognition-benchmark/blob/master/create_lmdb_dataset.py)). Can there be any suggestion what I am doing wrong ? 