NotImplementedError: Fast decoding only supports a single shard.

[kannadaraj]

Hi,

Thanks for this great repo. But this error after trying to use a newly trained model.
decoding command g2p-seq2seq --decode 1 --model_dir ./models --return_beams --beam_size 2
Error

python3.6/site-packages/tensor2tensor-1.7.0-py3.6.egg/tensor2tensor/models/transformer.py", line 497, in _fast_decode
    raise NotImplementedError("Fast decoding only supports a single shard.")
NotImplementedError: Fast decoding only supports a single shard

Please can anyone help?