Chinese decoding garbled characters in stream mode

[fan-niu]

The character "聪" is decoded into two tokens

`data: {"batch_index":0,"context_logits":0.0,"cum_log_probs":0.0,"generation_logits":0.0,"is_final":0,"model_name":"ensemble","model_version":"1","output_log_probs":0.0,"sequence_end":false,"sequence_id":0,"sequence_start":false,"text_output":"你"}

data: {"batch_index":0,"context_logits":0.0,"cum_log_probs":0.0,"generation_logits":0.0,"is_final":0,"model_name":"ensemble","model_version":"1","output_log_probs":0.0,"sequence_end":false,"sequence_id":0,"sequence_start":false,"text_output":"有点"}

data: {"batch_index":0,"context_logits":0.0,"cum_log_probs":0.0,"generation_logits":0.0,"is_final":0,"model_name":"ensemble","model_version":"1","output_log_probs":0.0,"sequence_end":false,"sequence_id":0,"sequence_start":false,"text_output":"�"}

data: {"batch_index":0,"context_logits":0.0,"cum_log_probs":0.0,"generation_logits":0.0,"is_final":0,"model_name":"ensemble","model_version":"1","output_log_probs":0.0,"sequence_end":false,"sequence_id":0,"sequence_start":false,"text_output":"�"}

data: {"batch_index":0,"context_logits":0.0,"cum_log_probs":0.0,"generation_logits":0.0,"is_final":0,"model_name":"ensemble","model_version":"1","output_log_probs":0.0,"sequence_end":false,"sequence_id":0,"sequence_start":false,"text_output":"明"}
`

Anyone can help to deal with this problem? thanks so much