aky15

I mean the usage of left_chunk_size is somewhat inconsistent with the definition in [the tutorial](https://github.com/espnet/espnet/blob/master/doc/espnet2_tutorial.md) and [the encoder](https://github.com/espnet/espnet/blob/master/espnet2/asr_transducer/encoder/building.py), which is 'the number of frames in left context'. If left_chunk_size and...

The left_context and right_context in inference are used in the same manner as num_left_chunks and num_right_chunks, but denoted as the num of frames in left/right context in the tutorial.

OK, Thanks for your contribution!

可以报一下数据集上的结果看看实验是否正常